>
Pipelines, quality, lineage and ownership: the engineering that decides which AI programmes compound.
Volge Quarterly is Volge's signature series, published every quarter for enterprise customers. Each edition takes one subject, sets out where it currently stands on the published record, and makes a set of dated, falsifiable judgements on where it is going, written to give a buyer something firmer than a forecast: every figure carries its source and its date, every judgement carries the date on which it resolves, and the following year's edition grades each one hit, missed or partial.
Plumbing before intelligence. The Gulf is buying artificial intelligence faster than it is building the data foundations beneath it; 2027 is the year data engineering becomes the binding constraint.
Figures carry their basis and their date. Where published estimates disagree, the range is given and the basis named rather than averaged away, and where a widely quoted number does not survive checking it is corrected here rather than repeated. Every chart plots a figure stated in the text, and nothing is projected beyond what a named source projects.
2027 is the year the Gulf's data foundations, not its models or its capital, decide what artificial intelligence delivers.
The Gulf is buying intelligence faster than it is building the plumbing beneath it. The order of the two matters, because the plumbing is what the intelligence runs on. Compute has been secured at national scale, capital is effectively unconstrained, and the political will to lead in artificial intelligence is stated at the highest level of government. What has not kept pace is the unglamorous layer between the raw data and the model: the ingestion, the cleaning, the cataloguing, the governance and the integration that turn scattered records into a foundation an algorithm can stand on. In 2027 that layer becomes the binding constraint, and the edition's single thesis is that the constraint is now made of data engineering rather than of chips, money or ambition.
The evidence for the ceiling is by now hard to dispute. A preliminary MIT Project NANDA study found that roughly 95 per cent of organisations in its evidence base, 52 interviews and 153 survey responses, had recorded no measurable profit-and-loss impact from generative AI [MIT Project NANDA, July 2025]. The finding is about organisations rather than about individual pilots, and it is the ceiling on measurable return, an outcome attributed to weak data foundations rather than to model quality [MIT NANDA, "State of AI in Business 2025", via Forbes, 15 October 2025]. The finding is not that the models are inadequate; it is that they are being asked to reason over data that has not been engineered to be reasoned over. Gartner has reached a compatible conclusion, observing that half of generative-AI projects fail and that more than 40 per cent of agentic-AI projects are expected to be cancelled before the end of 2027 [Gartner, 26 January 2026; 25 June 2025]. The pattern is consistent across the region's larger economies, where capital expenditure has been observed to race ahead of data and governance maturity [McKinsey, "The state of AI in GCC countries", 13 November 2025].
This is not a pessimistic read. The market is large and its trajectory is steep: the GCC cloud market was worth in excess of thirty billion US dollars in 2025, with Saudi Arabia at 15.39 billion and the United Arab Emirates at
12.84 billion, the latter growing at a compound annual rate near 28 per cent to 2031 [Mordor Intelligence, 2025]. The regional data-centre market has been sized at 2.39 billion US dollars in 2025, projected to reach 21.67 billion by 2034 [IMARC, 2025]. The buildout is real, it is well funded, and it is accelerating. What is argued here is narrower and more useful: that the return on all of it turns, in 2027, on a layer that has been treated as an afterthought.
The quieter story is where advantage now moves. It moves away from the entities that have bought the most compute and towards those that have engineered the cleanest, best-governed and most interoperable data; and it moves from the system integrator, who builds and departs, to the in-house platform team, who owns and compounds. The Gulf has proven it can buy anything it chooses. Whether it can build the foundation is the question of the year, and it is answered one data platform at a time.
The sequence is the argument: foundations first, or the intelligence layer has nothing sound to stand on.
Three facts, taken together, define the year. The first is that power and land are cheap and abundant: industrial electricity in Riyadh and Abu Dhabi runs at roughly 0.05 to 0.07 US dollars per kilowatt-hour, some five to seven times below the United Kingdom [Oxford Economics, 22 January 2026]. The second is that capital is deep beyond ordinary constraint: the Public Investment Fund reports assets of 900 billion US dollars and above, about SAR 3.4 trillion at end-2025 [PIF, 2025 annual report]; third-party rankings put it near 1.15 trillion by June 2026, sitting alongside Mubadala at roughly 330 billion and dedicated vehicles such as MGX and the HUMAIN programme [sovereign disclosures, 2025–26]. The third is that the data foundation is thin: the cost of poor data quality has been estimated at about 12.9 million US dollars per organisation each year [Gartner, via IBM, 23 January 2026].
The consequence is precise. Where compute and capital are abundant and the foundation is scarce, the scarce input sets the return. It is to be noted that this is the reverse of the position most Gulf entities believe themselves to be in; the perceived constraint has been access to chips and to models, and the actual constraint has quietly become the ability to feed them. PwC has projected that artificial intelligence will add roughly 320 billion US dollars to the Middle East by 2030, with Saudi Arabia at about 135 billion and the UAE at about 96 billion [PwC, 2023–24]; the realisation of that figure is now a data-engineering question rather than a compute one.
The Gulf has secured the compute, the power and the capital for artificial intelligence. In 2027 the return on all three is decided by the data-engineering layer beneath them, which has been built last and remains the region's binding constraint.
What follows reads that layer end to end. Part I establishes the state of the sector; Part II isolates the five forces; Part III makes the Calls; Part IV reads the year by domain and runs the scenarios; Part V is the toolkit and the annex. Throughout, the register is deliberately plain and every figure is sourced and year-stamped, so that an operator, an investor or a board may see where the region stands and where it is going.
The editors, Volge · Panache Editions, July 2026.
Two points qualify the totals. The capacity figure is a live-load measure and not a measure of what that capacity is made to do: Middle East live IT capacity was 761 megawatts in 2024, 1,115 in 2025 and is forecast at 1,669 in 2026, with the UAE holding 35.2 per cent of the GCC and Khazna alone carrying a pipeline above 300 megawatts [Knight Frank / DC Byte; IMARC / Business Wire, January 2026]. The colocation segment across the Middle East and Africa has been sized at 3.33 billion US dollars in 2025, rising to 12.88 billion by 2031 at a compound rate of 25.32 per cent, at a build cost of six to twelve million US dollars per megawatt [Arizton, 2025]. The megawatts are being poured; the question the edition presses is what fraction sits on data that has been engineered to be used.
Reconciled from Mordor Intelligence and IMARC (cloud and data-centre sizing, 2025), Knight Frank / DC Byte (live capacity, 2024–26), sovereign disclosures (PIF, 2025–26), PwC (AI value, 2023–24) and Oxford Economics (power, January 2026). Ranges reflect genuine source disagreement on scope. Current to July 2026.
A cloud market above thirty billion US dollars in 2025, a data-centre buildout compounding near 27 per cent, and five national data programmes racing to govern it, set against a data-engineering layer that has not kept pace.
The Gulf's data infrastructure is large, well funded and compounding fast; the figures below are the reconciled basis for the edition.
The scale of the buildout is not in dispute; the disagreement among sources is one of scope and definition rather than of direction. The GCC cloud market stood above thirty billion US dollars in 2025 on a national build-up, led by Saudi Arabia at 15.39 billion, forecast to 27.93 billion by 2030 at a compound annual rate of 12.66 per cent, and the United Arab Emirates at 12.84 billion, forecast to 56.26 billion by 2031 at 27.93 per cent [Mordor Intelligence, 2025]. The regional data-centre market has been sized at 2.39 billion US dollars in 2025, projected to 21.67 billion by 2034 at a compound rate of 26.78 per cent [IMARC, 2025].
Every Gulf state has stood up a national data authority; the maturity of what sits beneath the authority varies widely.
Saudi Arabia has gone furthest in institutional terms. The Saudi Data and Artificial Intelligence Authority was established by Royal Order A/471 on 30 August 2019, reports to the Prime Minister, and is overseen by a board chaired by the Crown Prince, with Dr Abdullah Alghamdi as president; it houses the National Information Centre, the National Data Management Office and the National Centre for AI [SDAIA, 2019]. The National
Strategy for Data and AI was approved on 17 July 2020, targeting more than 20,000 data and AI specialists and around 75 billion Saudi riyals of investment, with global leadership stated as the 2030 goal [NSDAI, 2020]. The National Data Management Office framework is unusually concrete: fifteen domains, seventy-seven controls and 191 specifications, arranged as a hierarchy from domain to control to specification, with a data-classification register mandated under its thirteenth domain [NDMO Standards, 2022].
The scale of integration already achieved is significant. The National Data Bank and associated lake integrate more than 360 government systems; the National Data Index covers more than 370 systems across in excess of 130 Government Data Offices; and the open-data portal data.gov.sa is the public face of the estate [SDAIA, 2025]. Adoption of the national digital-identity rails is high, with Nafath at roughly 21.9 million active users and Tawakkalna at around 33 million [SDAIA, 2025]. Data is classified across four levels (Top Secret, Secret, Confidential or Restricted, and Public), which is the mechanism through which residency and access are subsequently governed.
The United Arab Emirates governs federal data through policy rather than through a single statute. The Smart Data Strategy and its accompanying Smart Data Framework (Part 1 on principles and standards, Part 2 on implementation, drawing on the European Interoperability Framework, ISO and W3C) set the federal standard, with bayanat.ae operating as the federal open-data portal under the Federal Competitiveness and Statistics Centre [FCSC, 2025]. It is to be noted that this portal is distinct from the G42-linked commercial entity of a similar name. The federal Data Index measures entity maturity across regulatory (30 per cent), quality and protection (30 per cent) and accessibility and flow (40 per cent) dimensions. The Personal Data Protection Law, Federal Decree-Law 45 of 2021, has been in force since 2 January 2022, though its executive regulations remained unpublished in mid-2026, a material gap for any entity engineering for compliance.
Below the federal layer, Dubai and Abu Dhabi run the region's most advanced sub-national data platforms, and Qatar has moved quickly to catch up.
Dubai treats data as an owned asset in law. Dubai Data Law No. 26 of 2015 established that Dubai Data is owned by government and required all such data to be classified as Open or Shared; the governing body has since become the Dubai Data and Statistics Establishment under Law 24 of 2023, sitting within the Dubai Digital Authority created by Law 22 of 2023 [Government of Dubai, 2015–23]. The central platform, Dubai Pulse, delivers platform, data and infrastructure as a service, built with the operator du. An updated Dubai Data Manual was launched on 1 July 2026, adding mandatory standards, self-assessment checklists, an AI-readiness scope and a Knowledge Hub on Data.Dubai [DDSE, 1 July 2026]. The Dubai Universal Blueprint for AI, launched on 29 April 2024 with 22 Chief AI Officers, supports the D33 target of 100 billion dirhams a year from digital.
Abu Dhabi has consolidated its data function within the Department of Government Enablement, established in 2023, which absorbed the former Abu Dhabi Digital Authority. The emirate's Government Digital Strategy 2025–2027 commits 13 billion dirhams, targets full sovereign-cloud adoption for government, and states the ambition of the world's first fully AI-native government by 2027, with a projected contribution above 24 billion dirhams to GDP and more than 5,000 jobs [DGE, 2025]. In a concrete step, the Department and e& announced on 17 October 2025 an Abu Dhabi Unified Government Data Centre to power more than forty government entities, precisely the kind of consolidation that the platform thesis of this edition anticipates.
Qatar has moved from a standing start with speed. The National Planning Council was established by Amiri Decision 13 of 2024; the National Data and Statistics Strategy was launched on 4 May 2025; and the National Data Program, launched in January 2025, provides a central national data repository, a National Data Policy and Standards, a Qatar Data Trust Index and the QInsights platform of 119 indicators [NPC, 2024–25]. Taken together, the five programmes show a region that has largely solved the question of who governs data and has not yet solved the question of
whether the data beneath the governance is engineered to be used.
The physical foundation is a lattice of hyperscaler regions and sovereign stacks, most of it stood up since 2019 and much of it arriving in 2026.
The estate on which Gulf data platforms are built is young. The first hyperscaler region in the Middle East, AWS Bahrain, reached general availability only in July 2019; the densest wave of capacity arrives in 2026, when both the Microsoft Saudi Arabia East region and the AWS Saudi region are due. The table sets out the principal hyperscaler footprint with dates, which is the single most useful reference for where a compliant, in-region data platform can actually be built today.
| Provider | Region | Status & date | Commitment |
|---|---|---|---|
| Microsoft Azure | UAE North / Central | Live 2019 | UAE spend >$15.2B by end-2029; G42 deal $1.5B (Apr 2024) |
| Microsoft Azure | Qatar Central | GA Aug 2022 | Cloud for Sovereignty GA Dec 2023 |
| Microsoft Azure | Saudi Arabia East (Riyadh) | Customer availability targeted Nov 2026 (3 AZs) | Construction complete on 3 sites, Dec 2024 |
| AWS | Bahrain (me-south-1) | GA Jul 2019 (first ME, 3 AZs) | – |
| AWS | UAE (me-central-1) | GA Aug 2022 | ~$5B by 2036 |
| AWS | KSA (me-central-2) | Announced Mar 2024, launch 2026 | $5.3B |
| Google Cloud | Doha, Qatar | Live 2023 | – |
| Google Cloud | Dammam, KSA | Live 2024 | JV with Center3 / stc; no UAE region |
| Oracle | UAE & KSA (Jeddah, Riyadh) | Jeddah live ~2020 | $14B KSA over 10 years (May 2025) |
Sources: provider announcements and general-availability notices, 2019–2026; Microsoft–G42 (April 2024); Oracle KSA commitment (May 2025). Current to July 2026.
Beneath the hyperscalers sits a sovereign layer that is where much government data will in practice reside. In Saudi Arabia this includes Center3, backed by stc and targeting one gigawatt by 2030 on a ten-billion-dollar plan of which roughly three billion has been spent, alongside Alat, backed by the Public Investment Fund, and SDAIA's own capacity [company disclosures, 2025]. In the UAE it includes Core42 and G42's sovereign cloud running on Azure, e& enterprise, du, Khazna and Moro Hub; in Qatar, Ooredoo and Meeza; in Bahrain, Beyon; and in Oman, Oman Data Park. The estate is therefore hybrid by default, and it is this hybridity (data spread across hyperscaler regions and sovereign stacks, under differing classification regimes) that makes the engineering of a single, governed data platform hard.
Read country by country, the region divides into a leading tier that has institutionalised data governance and a following tier still assembling it.
Maturity in data engineering is not the same as ambition in artificial intelligence, and the two are frequently confused. The assessment below rates each state on the depth of its national data programme, the integration achieved across government systems, and the readiness of its foundation to support production AI, as distinct from the scale of compute it has procured. It is to be noted that the ordering tracks institutional depth rather than headline spending.
| Country | Governing body | Foundation depth | AI-readiness of data |
|---|---|---|---|
| Saudi Arabia | SDAIA (NIC, NDMO, NCAI) | High: 360+ systems integrated; 15 domains, 77 controls, 191 specifications | Advancing: strong governance, engineering talent the constraint |
| UAE (federal) | FCSC / Smart Data Framework | Medium-High: policy-led; PDPL exec regs unpublished mid-2026 | Advancing: measured by Data Index; uneven across entities |
| Dubai | DDSE / Dubai Digital Authority | High: data-as-asset since 2015; Pulse platform; Manual updated Jul 2026 | Advancing: AI-readiness now in scope of the Data Manual |
| Abu Dhabi | DGE (absorbed ADDA, 2023) | Medium-High: unified data centre for 40+ entities (Oct 2025) | Ambitious: "AI-native government" targeted for 2027 |
| Qatar | National Planning Council | Medium: National Data Program stood up Jan 2025 | Early: repository and standards new; QInsights live |
| Bahrain | iGA / Beyon | Medium: first-mover on cloud (AWS 2019), smaller estate | Early: narrower programme, limited disclosure |
| Oman | MTCIT / Oman Data Park | Low-Medium: nascent national programme | Early: foundation-building stage |
Volge assessment, July 2026, drawing on the national-programme sources cited in section 02. Ratings are relative within the region and reflect data-foundation depth, not compute procurement or AI ambition.
The practical implication is that the leading tier (Saudi Arabia, Dubai and Abu Dhabi) has the governance scaffolding largely in place and now faces an execution problem in engineering and talent, while the following tier still faces an institutional one. For both, the constraint of 2027 sits at the level of engineering capacity rather than of policy.
The foundation gap is measurable, and the clearest measure is the rate at which generative-AI initiatives fail for want of sound data.
The single most important figure in this edition is that roughly 95 per cent of organisations in MIT Project NANDA's evidence base recorded no measurable profit-and-loss impact from generative AI [July 2025], an outcome attributed to weak data foundations rather than to model quality [MIT NANDA, "State of AI in Business 2025", via Forbes, 15 October 2025]. The figure is not a comment on the sophistication of the models being deployed; it is a comment on the state of the data being fed to them. Where the data is fragmented, poorly catalogued, inconsistently classified and ungoverned, no model, however capable, can produce a reliable result, and the pilot stalls before it reaches production.
MIT NANDA / Forbes (15 October 2025); Gartner (26 January 2026; 7 April 2026; 25 June 2025); McKinsey (13 November 2025); Gartner via IBM (23 January 2026); Harness (2025); Palo Alto Networks, 2026 Identity Security Landscape. Two developments postdate the research and bear on every Saudi and UAE entry: SDAIA consulted from 6 October to 5 November 2026 on draft amendments that would require personal data to be kept in the Kingdom, and the UAE established a Federal Authority for Artificial Intelligence and Data on 14
June 2026, which absorbs the Data Office and is the expected issuer of the PDPL executive regulations. Current to November 2026.
The corroborating evidence is consistent. Gartner has separately observed that AI infrastructure-and-operations projects stall before reaching return on investment, and that more than 40 per cent of agentic-AI projects will be cancelled by the end of 2027 [Gartner, 7 April 2026; 25 June 2025]. McKinsey, examining the Gulf specifically, has found capital expenditure racing ahead of data and governance maturity [McKinsey, 13 November 2025]. The gaps beneath these figures are well understood by practitioners: a shortage of data-engineering talent; over-reliance on system integrators; governance treated as an afterthought; tool sprawl; migration and technical debt; and, above all, data quality inadequate for grounding an AI model. Each is an engineering problem, and each is why the foundation, not the model, is the ceiling.
Five through-lines cut across the region. Each is a force that will move data engineering in 2027, and together they explain the Calls that follow.
Classification-driven data residency is moving from a compliance overlay to a first principle of architecture.
Sovereignty in the Gulf is ceasing to be a policy stated after a system is built and is becoming a constraint designed into it from the first line. The mechanism is data classification: Saudi Arabia's four-level scheme (Top Secret, Secret, Confidential or Restricted, and Public) and the parallel schemes in the UAE, Dubai and Qatar determine not merely who may see a record but where, physically and jurisdictionally, that record may be processed and stored [SDAIA; DDSE; FCSC, 2015–26]. When residency is a function of classification, and classification is mandated at the level of the individual data element, sovereignty becomes an engineering property of the platform rather than a legal wrapper around it.
The practical consequence for data engineering is significant. A platform built for classification-driven residency must carry the classification with the data through every pipeline, enforce it at query time, and
prove it under audit, which demands catalogue, lineage and policy engines of a maturity that most Gulf estates do not yet possess. The Dubai Data Manual's addition of AI-readiness scope and self-assessment checklists on 1 July 2026 is an acknowledgement that these controls must now be demonstrable rather than asserted [DDSE, 1 July 2026].
The evidence that vendors are engineering for this is direct. Microsoft's Cloud for Sovereignty reached general availability in December 2023, and its 1.5-billion-dollar arrangement with G42 in April 2024 was explicitly framed around sovereign controls; Oracle offers Alloy, Dedicated Region and Sovereign Cloud precisely to place a compliant boundary around regulated data [provider disclosures, 2023–25]. Abu Dhabi has gone further in policy, targeting 100 per cent sovereign-cloud adoption for government within its 2025–2027 strategy [DGE, 2025]. Sovereignty is thus not a niche requirement but the default posture of the region's largest data owner, its government.
It is to be noted that this force is, on balance, constructive. Classification-driven design forces the cataloguing, lineage and governance work that a sound data foundation requires in any case, and which is too often deferred. The entities that treat sovereignty as an architectural discipline rather than a documentation exercise will, as a by-product, build the cleaner and better-governed foundation that production AI depends upon.
The scarcest input in the Gulf data estate is not compute or capital but the engineers who can build and run the platforms.
The binding constraint on Gulf data engineering in 2027 is human, not physical. Capital is effectively unconstrained and compute is being procured at national scale, but the supply of data engineers capable of designing pipelines, governing catalogues and operating platforms is thin relative to demand, and it cannot be expanded at the speed capital can be deployed. Saudi Arabia's National Strategy for Data and AI itself targets the training of more than 20,000 data and AI specialists by 2030, a figure that is both an
ambition and an admission of the size of the present shortfall [NSDAI, 2020].
The gap is structural rather than cyclical. Localisation programmes across the Gulf are producing graduates, but the specific discipline of data engineering (as opposed to data science or general software development) has a long apprenticeship, and the regional pool of senior practitioners who have built and run production platforms at scale remains small. The consequence has been a heavy reliance on expatriate talent and on system integrators to supply capability that the entity does not hold in-house, which defers rather than resolves the shortage.
The scale of the estate that this thin talent base must govern is easily underestimated. There are now roughly 109 machine identities for every human in a modern enterprise [Palo Alto Networks, 2025], each generating data that must be ingested, secured and governed; and the cost of failing to do so has been estimated at about 12.9 million US dollars per organisation each year in poor data quality alone [Gartner, via IBM, 23 January 2026]. The mismatch between the size of the estate and the size of the workforce able to engineer it is the mechanism by which the talent gap becomes a foundation gap.
This is the force least amenable to a fast solution, and it is therefore the one held with the highest confidence in the Calls that follow. Capital can buy compute in a quarter and a data platform in a year; it cannot manufacture a senior data engineer in either. Through 2027 the demand for data-engineering talent is expected to continue outstripping locally available supply despite localisation, and the entities that succeed will be those that build durable in-house teams rather than renting capability by the project.
The operating model is shifting from the one-off, integrator-led build towards the durable, in-house data platform.
The dominant way Gulf data capability has been delivered (as a discrete project, scoped, built and handed over by a system integrator) is reaching the limit of its usefulness. A project ends; a platform compounds. As entities discover that the value of their data lies in its continuous engineering rather than in a single
delivered system, the operating model is beginning to shift towards owned platforms staffed by in-house teams, with the integrator moved from builder to partner. It is to be noted that this shift is early and uneven, but its direction is clear.
The structural driver is that AI-grade data foundations are not a state to be reached and left, but a capability to be operated. Catalogues drift, classifications change, sources multiply, and models require continuously curated, high-quality data to remain reliable. An estate that has 109 machine identities per human and integrates hundreds of source systems (Saudi Arabia's National Data Bank alone spans more than 360) cannot be maintained by a vendor engagement that concludes on handover [Palo Alto Networks, 2026; SDAIA, 2025].
The consolidation now visible in the physical layer prefigures the same move in the logical one. The Abu Dhabi Unified Government Data Centre, announced by the Department of Government Enablement and e& on 17 October 2025 to power more than forty government entities, is a platform play rather than a project: shared infrastructure, operated centrally, serving many consumers [DGE, 2025]. The same logic (build once, operate continuously, serve many) is what distinguishes a data platform from a data project, and it is the logic the region is slowly adopting.
The barrier is the talent bottleneck of the preceding force. A platform operating model requires a standing team, and standing teams are precisely what the region is short of. The Calls therefore treat this shift as real but gradual, expecting it to appear first at a small number of flagship entities in 2027 rather than across the estate.
The artificial-intelligence layer the region is buying will only perform on a data foundation engineered to feed it.
The central irony of the Gulf's position is that the more it invests in artificial intelligence, the more exposed it becomes to the weakness of its data foundation. A model is only as good as the data it is grounded in, and the failure rate of generative-AI initiatives (roughly 95 per cent of organisations recording no measurable profit-and-loss impact, attributed to
data foundations rather than to model quality) is the clearest possible statement that the constraint has moved downstream of the model [MIT NANDA, via Forbes, 15 October 2025]. Building for AI, correctly understood, is therefore an instruction to build the data foundation first.
The corroboration is broad and recent. Gartner has found that about half of generative-AI projects fail, that AI infrastructure-and-operations projects stall before reaching return on investment, and that more than 40 per cent of agentic-AI projects will be cancelled by the end of 2027 [Gartner, 26 January 2026; 7 April 2026; 25 June 2025]. McKinsey, examining the Gulf, has found capital expenditure racing ahead of data and governance maturity [McKinsey, 13 November 2025]. None of these findings implicates the models; all implicate the foundation.
The engineering requirement that "building for AI" imposes is exacting. An AI-ready foundation demands high data quality, complete and current metadata, enforced classification, traceable lineage, and the retrieval infrastructure to serve grounded context to a model at query time. These are the disciplines of data engineering, and they are the disciplines the region has under-invested in relative to compute. The addition of an explicit AI-readiness scope to the Dubai Data Manual on 1 July 2026 is a sign that the more advanced entities now recognise this [DDSE, 1 July 2026].
The implication for 2027 is direct and appears in the Calls. A majority of the enterprise generative-AI initiatives launched across the Gulf in 2025 and 2026 are expected to stall or be re-scoped pending foundation work, and at least one national entity is expected to report publicly an AI failure or rescope explicitly attributed to data quality or engineering. Both are the predictable consequence of loading intelligence onto a foundation that was built last.
As cloud-and-data spend outpaces value, financial discipline arrives for the data estate as a named function.
The era of unmetered data spending in the Gulf is ending. With capital abundant and compute cheap, data and cloud budgets have grown with little scrutiny of the value returned; in 2027 that scrutiny
arrives. The trigger is the visible gap between spend and value (projected global cloud waste of 44.5 billion US dollars in 2025, with Flexera putting waste at 29 per cent of global infrastructure and platform spend in 2026, the highest in five years, and an estimated 12.9 million US dollars a year in poor-data-quality cost per organisation), which is now large enough that finance functions can no longer treat data as an unquestioned overhead [Harness, 2025; Gartner via IBM, 23 January 2026].
The discipline has a name: FinOps, applied to data. It treats the data estate as a cost centre to be measured, attributed and optimised: storage tiered to access, pipelines rationalised, duplicate tooling retired, and spend attributed to the business unit that incurs it. The mechanism matters because the Gulf's cost advantage in raw inputs, with industrial power at 0.05 to 0.07 US dollars per kilowatt-hour, has masked inefficiency in how those inputs are consumed [Oxford Economics, 22 January 2026]. Cheap power makes waste affordable, which is precisely why the waste has accumulated.
Tool sprawl is the clearest target. The proliferation of overlapping ingestion, catalogue, quality and observability tools (each procured for a project, none retired when the project ended) is both a cost and an engineering liability, because every additional tool is another surface to govern across an estate already carrying 109 machine identities per human [Palo Alto Networks, 2026]. Consolidation onto a smaller, governed platform stack reduces spend and improves the foundation simultaneously, which aligns the financial and engineering cases.
The Call that follows expects FinOps-for-data to become a named function or explicit line item at major Gulf entities during 2027, as the gap between data spend and data value becomes too large to leave unmanaged. It is the mechanism by which cost discipline, arriving late, begins to force the foundation work that abundance had allowed the region to defer.
Eight rows for six states: the United Arab Emirates governs data at three levels, and the federal instrument, the Abu Dhabi policy and the Dubai policies document impose different duties on different entities.
What the rules require of data itself, where the region actually stands, what the platforms cost in the Gulf, and which maturity model a buyer can use without a licence.
Three of the six GCC states mandate a data catalogue in an instrument a buyer can read. None of the six mandates data lineage, and no state has published a penalty under a data-management instrument as distinct from a privacy law.
The first question to settle in a data programme is not what good practice recommends but what the jurisdiction compels. On that test the Gulf position is considerably narrower than the market assumes. Three of the six GCC states mandate a data catalogue in an instrument that can be read in full: Abu Dhabi, Oman and Kuwait. Not one of the six mandates data lineage. And no GCC state has published a penalty attached specifically to a data-management or data-classification instrument, as opposed to a privacy law. The table below sets out the position state by state, against the instrument, what is mandated, who is bound and how compliance is assessed. It is confined to what was readable. Where an obligation rests on vendor restatement rather than on an operative clause, that is said in terms.
| State | Instrument | What is mandated | Who is bound | Assessment regime |
|---|---|---|---|---|
| Saudi Arabia | National Data Governance Interim Regulations, SDAIA/NDMO; Data Management and Personal Data Protection Standards, NDMO, v1.0 June 2020 | A Data Management Office, a Data Management Committee and a Chief Data Officer at each entity. Catalogue, retention and master data appear in vendor restatements, not verified against the instrument | All Kingdom public entities, extending to business partners handling government data (vendor source) | Annual compliance audit, report submitted to NDMO in Q3 each year; ad-hoc audits possible; NDMO states it reports at entity, sector and whole-of-government level |
| UAE, federal | Federal Decree-Law 44 of 2021, in force 2 January 2022 | Creation of the Data Office and the empowering of a regulator. No catalogue, quality or lineage duty of its own | Federal level | None stated in the instrument |
| State | Instrument | What is mandated | Who is bound | Assessment regime |
|---|---|---|---|---|
| Abu Dhabi | Government Data Management Policy v1.0, ADSIC-DM-PD03; issue date not public | Thirteen domains. A bilingual Data Catalogue, mandatory metadata keys, data quality metrics, lifecycle and destruction controls, reference and master data toolsets, a single enterprise data model | Abu Dhabi government entities, which must bind their suppliers | Programme status reported against Data Management Programme Plan milestones, frequency stated only as regular; central assessment of programmes and technical controls |
| Dubai | Law 26 of 2015 and Resolution 2 of 2017 Policies Document | Classification before dissemination or exchange (Article 7). Open Data and Shared Data, the latter split Confidential, Sensitive and Secret | Dubai government entities | No assessment regime found; no chief data officer mandate found; no penalty found |
| Qatar | National Data Policy, October 2024; National Data Management Guidelines, July 2025; National Data Classification Policy v3.0, launched 7 May 2023 | Reported as five layers and twelve domains binding entities to national standards. Guidelines published per domain, including Data Catalog and Metadata Management; mandatory or advisory not established. Five classification tiers, C0 Public to C4 Top Secret | All government entities; the classification policy also binds vital-sector entities, with a minimum of four levels for non-government | Internal review against the National Data Standards and Qatar Data Trust Index scoring; cycle not dated |
| Bahrain | None issued. The iGA AI Readiness Assessment Methodology report of 12 November 2025 describes a National Data Governance Framework as still in development | No catalogue, quality, chief data officer or master data mandate found | Not applicable | None |
| State | Instrument | What is mandated | Who is bound | Assessment regime |
|---|---|---|---|---|
| Oman | Regulatory Framework for National Data Governance and Management, Ministerial Circular 113/2025, MTCIT; policies document dated 1 January 2025 | Thirteen policies, including Data Catalogue and metadata, Data Classification, Data Quality, Reference and Master Data, Open Data and Data Monetisation. A Data Governance and Management Office with named roles | All government units; the private sector reached indirectly through sector regulators | Annual periodic assessment by the ministry's compliance team; provisions split across three years, Priority 1 in year one, Priority 2 in year two, Priority 3 in year three |
| Kuwait | Data Classification Policy v2.3, CITRA; no issue date and no commencement date on the document | The data owner must create and maintain a data catalogue including metadata in a unified format, updated periodically (clause 3.1.3). Encryption of Tiers 3 and 4 in transit, a classification team headed by senior management, a classification road map | Public and private sector entities. The only GCC data-classification instrument binding private business directly | Quarterly implementation reports to CAIT; CITRA periodic compliance reports must contain the full catalogue with metadata, approved tiers with criteria, storage locations by tier and progress against the road map (clause 3.4.2.3) |
What the table shows is a region legislating at three quite different levels of specificity. At one end sit Abu Dhabi, Oman and Kuwait, whose instruments can be read clause by clause and whose obligations can therefore be costed. At the other sits Bahrain, which has issued nothing, and the UAE federal layer, where Decree-Law 44 of 2021 is institutional rather than operational: it creates a regulator and empowers it, and imposes no duty on an entity's data estate at all. Between the two sit Saudi Arabia and Qatar, both of which have built substantial apparatus whose operative text is not reachable. The position is that the weight of a state's data regime and the legibility of it are not correlated, and the two states with the largest programmes are the two about which least can be stated with confidence.
On cataloguing, the three mandates differ in character. Kuwait's is the bluntest: clause 3.1.3 of the CITRA policy puts the duty on the data owner to create and maintain a catalogue including metadata in a unified format, and clause 3.4.2.3 requires the catalogue itself to be surrendered in periodic compliance reporting. Abu Dhabi's is the most demanding: a bilingual catalogue covering ownership, provenance, composition and quality requirements, which must be checked before any new data definition is created. Oman's sits within a numbered policy on Data Catalogue and metadata inside a thirteen-policy framework. Saudi Arabia is widely described as mandating a catalogue, but the obligation appears in vendor restatements rather than in any NDMO text that could be read, and that distinction matters when a buyer is asked to justify a budget.
Data lineage, the recorded path by which a figure in a report traces back to the system that produced it, is not an express obligation anywhere in the GCC. It was not found in any readable instrument in any of the six states. Abu Dhabi comes closest: its catalogue requirement extends to provenance, which is adjacent to lineage without being it. The consequence is worth stating plainly, because it cuts against the direction of the market. Lineage is the control that answers the question an auditor, a regulator or a board will eventually ask of an AI output, which is where the number came from. No Gulf state currently compels it. A buyer who defers it until it is compelled will be building it under time pressure, retrospectively, across systems already in production.
No GCC state publishes a penalty attached specifically to a data-management or data-classification instrument. Oman publishes no penalty under Circular 113/2025. Dubai publishes none under Resolution 2 of 2017. The Abu Dhabi policy is mandatory and carries no published sanction. The contrast with privacy law is sharp. Saudi Arabia's Personal Data Protection Law, Royal Decree M/19 of 2021 as amended by M/148 of 2023, carries fines to SAR 5m which may be doubled, and SDAIA announced 48 enforcement decisions in
the year to early 2026 (Latham and Watkins, 15 May 2026), although the decision texts and the respondents are not public. Bahrain's Law 30 of 2018, in force 1 August 2019, carries up to one year's imprisonment and fines to BHD 20,000. The asymmetry is instructive: the Gulf has chosen to put its sanctions behind personal data and its ambitions behind data management.
It is to be noted that these instruments do not stop at the perimeter of government. Abu Dhabi requires entities to bind their suppliers, which places a contractor inside the policy whether or not the contract repeats its terms. Kuwait's classification policy binds private sector entities directly and is, on the evidence available, the only GCC data-classification instrument that does so. Oman reaches the private sector indirectly, through sector regulators. Saudi Arabia is reported to extend to business partners handling government data, though that reach too rests on a vendor source rather than on a clause. For a commercial buyer with public sector revenue in the Gulf, the practical reading is that the obligation arrives by contract before it arrives by statute.
The structure quoted almost universally for the Saudi NDMO Standards, fifteen domains, 77 controls and 191 specifications, appears only in vendor compliance material and in no reachable NDMO document. SDAIA's own National Data Index gives fourteen domains. The figures are treated here as vendor-originated. Similarly, the Saudi classification tiers conflict between two official SDAIA documents, one giving Top Secret, Secret, Confidential and Public, the other Top Secret, Secret, Restricted and Public. The two are unreconciled.
The takeaway for a buyer is that the compliance floor in the Gulf is a catalogue, a classification scheme and a governance body with a named accountable officer, and that nothing beyond that floor is currently compelled anywhere in the region. Everything above it, lineage included, is built for the buyer's own defensibility rather than for a regulator's. That is a better reason to build it, and a harder one to fund.
The Abu Dhabi Government Data Management Policy is the most demanding readable data instrument in the GCC. Its obligations pass down to suppliers, which means a vendor to an Abu Dhabi entity is inside the policy whether or not the contract repeats it.
If a buyer in the Gulf is to read one data instrument in full, it should be the Abu Dhabi Government Data Management Policy v1.0, document reference ADSIC-DM-PD03. It is the richest readable data instrument in the region: thirteen domains, mandatory in application, and specific enough that its requirements can be converted into a work plan without interpretation. Two features set it apart from everything else in the GCC.
The first is that its catalogue requirement operates as a design control rather than a documentation exercise. The second is that entities must bind their suppliers to it, which extends the policy well beyond the government payroll. One oddity should be recorded at the outset: the release date field in the version table of the document is blank, so the issue date of the most demanding data instrument in the Gulf is not public.
The policy begins where a buyer would expect an instrument of this kind to begin, with accountability, and then goes further than most. An entity must appoint a Data Manager who is accountable to the entity board. The seniority of that reporting line is the point: the policy does not permit data management to be housed as a technical function reporting into an information technology directorate. An entity must also constitute a Data Governance Board. And it must set key performance indicators for every one of the thirteen domains, which converts each domain from a statement of principle into something measurable. The practical effect is that an entity cannot claim partial adoption. Thirteen domains with indicators attached to each produce thirteen separate reporting obligations, and central assessment covers both programmes and technical controls.
The catalogue requirement is the clause most likely to be underestimated. An entity must maintain a bilingual Data Catalogue covering ownership, provenance, composition and quality requirements, and it must populate mandatory metadata keys. Each of those four coverage requirements carries its own cost. Ownership requires a named accountable person for every data asset, which in most estates does not exist and cannot be inferred. Provenance requires a record of where the data came from. Composition requires a description of what is in it. Quality requirements require a statement of the standard the asset is meant to meet, which presupposes that someone has decided what that standard is. The bilingual requirement doubles the drafting work.
The clause that changes behaviour, however, is the requirement that the catalogue be checked before any new data definition is created. That is not a reporting duty. It is a gate in the development lifecycle. An entity complying with it cannot stand up a new table, a new field or a new extract without first establishing that the concept does not already exist in the catalogue under another name. It is the only provision in any readable GCC instrument that attacks the actual mechanism by which data estates become unmanageable, which is uncontrolled proliferation of near-duplicate definitions.
On quality the policy requires data quality metrics supported by a prioritised profiling and cleansing plan. The word prioritised is doing work there: the policy does not require an entity to clean everything, which would be unachievable, but it does require the entity to have decided and recorded what it will clean first. On lifecycle, an entity must document lifecycle attributes including archive and destruction, and the destruction provision introduces a third party into what is otherwise an internal process: written approval from the National Archives is required before destruction. A retention schedule built on the assumption that an entity may delete its own data at the end of a defined period will not comply.
Two further requirements carry significant implementation weight. An entity must stand up reference and master data toolsets with match-and-merge rules, which is a tooling obligation and not merely a policy one: match-and-merge implies a resolution engine and a documented rule set, with all the governance that attaches to deciding when two records describe the same customer or the same asset. And an entity must hold its enterprise data model in a single repository. The word single is the operative one. Most government estates of any age hold several partial models in several tools, maintained by several teams, and the consolidation of those into one repository is typically a longer exercise than the cataloguing itself.
The provision with the widest commercial reach is the shortest. The policy is mandatory and entities must bind their suppliers. The consequence is that a systems integrator, an analytics vendor, a managed service provider or a cloud hosting partner working for an Abu Dhabi government entity is operating inside an
instrument it may never have read. Reporting reinforces this: the central assessment covers programmes and technical controls, and the contractors sit inside both. A buyer on the supply side should therefore treat the Abu Dhabi policy as a tender requirement rather than a client-side compliance matter, because the catalogue, the metadata keys and the lifecycle documentation will in practice be produced by the supplier even where the obligation rests on the entity.
Three absences deserve recording, because they are as useful to a buyer as the obligations. First, the reporting frequency is stated only as regular: entities report programme status against Data Management Programme Plan milestones, and no interval is published. An entity cannot therefore plan its reporting calendar from the instrument. Second, no penalty is published. The policy is mandatory and carries no stated sanction, which is the regional pattern described in the preceding chapter. Third, data lineage is not required. Provenance in the catalogue is the nearest thing to it, and provenance records an origin rather than a path. The most demanding data instrument in the Gulf does not compel lineage, and nor does any other.
The takeaway is that the Abu Dhabi policy is the most useful document in the region for a buyer anywhere in the Gulf, including a buyer with no Abu Dhabi exposure at all. It is specific, it is readable, and it describes in operative language what the other five states describe in aspiration. Used as a template rather than as a compliance burden, it supplies the work plan that the Saudi and Qatari instruments do not make legible.
Saudi Arabia has built the heaviest data governance apparatus in the Gulf and publishes the least about its results. The National Data Index has existed since November 2023 and no result from it has been published for any year.
The Saudi position is a paradox that a buyer must plan around rather than resolve. The Kingdom has the most elaborate data governance machinery in the region: a dedicated authority, a national data management office, a bundled set of national policies, mandated governance bodies inside every public entity, an annual audit with a fixed quarter and a purpose-built national index to score the result. What it does not have is a published outcome. No National Data Index result has been made public for any year,
despite NDMO's stated commitment to publish, and the operative text of the standards against which entities are audited was not reachable. The consequence for a buyer is specific: compliance in Saudi Arabia must be planned against an apparatus whose requirements are known largely by restatement and whose scoring is invisible.
Two instruments carry the regime. The National Data Governance Interim Regulations, issued by SDAIA and NDMO, appear as v1 dated 1 June 2020 on the official document, while secondary analysis gives approval on 28 September 2020 and publication on 20 October 2020. Those datings are unreconciled, and separate commencement dates for the component policies were not found. The Regulations bundle five policies: Data Classification, Data Sharing, Freedom of Information, Open Data and the Personal Data Protection Interim Regulations. Alongside them sit the Data Management and Personal Data Protection Standards, NDMO, v1.0 of June 2020, issued under Cabinet Resolution 292 of 27/04/1441H. Whether a second version of those Standards exists could not be established. For an instrument binding every public entity in the largest economy in the Gulf, that is a notable gap in the public record.
Three organisational requirements are clear. Each public entity is to establish a Data Management Office, constitute a Data Management Committee and appoint a Chief Data Officer. These are the structural obligations and they are not in doubt. Beyond them the picture becomes dependent on secondary material. Catalogue, retention and master data duties are listed in vendor restatements of the Standards and were not verified against the instrument itself. The frequently quoted structure of fifteen domains, 77 controls and 191 specifications appears only in vendor compliance writing and in no NDMO document that could be reached, and SDAIA's own National Data Index is built on fourteen domains rather than fifteen. The scope is reported to extend beyond government to business partners handling government data, which, if correct, is commercially significant for every supplier to the Saudi public sector. That too rests on a vendor source.
Because the verbatim operative clauses of the NDMO Standards were not obtainable, the question of which of catalogue, lineage, chief data officer and retention are strictly mandatory rests, for all but the chief data officer, on vendor restatement. A buyer relying on those duties in a board paper should say so.
The assessment regime is the most concrete part of the Saudi apparatus and the part a buyer can plan against. Every public entity conducts an annual compliance audit and submits an Entity Compliance Report to NDMO in the third quarter of each year. No fixed day within the quarter is published. NDMO may additionally run ad-hoc audits, and it states that it reports at entity, sector and whole-of-government level. That is a recurring, binding and dated obligation in a region where almost nothing else is dated, and it is the single most useful fact in Saudi data compliance. An entity that treats Q3 as the planning anchor will be working to the only published rhythm the regime has.
The National Data Index was unveiled by SDAIA in November 2023 and is unusually well specified. It has three components: a maturity questionnaire of 42 questions across fourteen domains scored on a six-level scale running from Level 0, Absence of Capabilities, to Level 5, Pioneer; a compliance assessment scored per specification, at 100 per cent for full implementation and zero for partial; and a set of operational indicators. The binary treatment of partial compliance is worth noting, since it means an entity nine-tenths of the way through a control scores the same as one that has not started. Results have not been published for any year. The only entity-level outcome that could be found anywhere is the recognition of the Ministry of Health as an advanced entity on 26 December 2024, which carries no score.
The consequence is that no Saudi entity can benchmark itself, and no supplier can calibrate a proposal against the actual state of the client's estate. It also leaves a vacuum that other indices are drawn into filling. The Digital Government Authority's Digital Transformation Index reported an average of 87.14 per cent for 2024, up from 69.39 per cent in 2021, across an entity population reported as both 233 and 235. That figure is frequently cited as evidence of Saudi data maturity. It is not: it measures digital government, not data management, and it should not be offered as a proxy.
A further unresolved point affects anyone implementing classification in the Kingdom. Two official SDAIA documents give different classification schemes: one gives Top Secret, Secret, Confidential and Public, the other Top Secret, Secret, Restricted and Public. The third tier is therefore unsettled on official sources. The practical advice is to implement the structure and label the third tier in accordance with whichever document the contracting entity cites, and to record the discrepancy in the design documentation rather than resolve it silently.
The Personal Data Protection Law, Royal Decree M/19 of 2021 as amended by M/148 of 2023, came into force on 14 September 2023 with a grace period that ended on 14 September 2024. Fines run to SAR 5m and may be doubled. SDAIA announced 48 enforcement decisions in the year to early 2026 (Latham and Watkins, 15 May 2026), though neither the decision texts nor the respondents are public. The contrast with the data-management side of the same authority is the finding: the privacy instrument is enforced and the enforcement is unpublished, while the data-management instrument is audited annually and the audit is unpublished. On capability, the one Gulf figure on specialist supply is also Saudi: SDAIA has trained more than 779,000 citizens including 9,775 specialists and 260 scientists and experts, per the 2024 Vision 2030 report (SPA N2311819, 5 May 2025), against an announced NSDAI target of approximately 20,000 data and artificial intelligence specialists by 2030.
The takeaway is that a buyer in Saudi Arabia should build to the Q3 audit, document every obligation that rests on restatement as resting on restatement, and resist the temptation to import the Digital Transformation Index average as a measure of data readiness. The apparatus is real. The score is not available, to anyone.
Oman and Kuwait have produced the Gulf's two most usable data instruments. Oman tiers its thirteen policies across three years by priority, and Kuwait's classification policy is the only one in the region that binds private business directly.
The two smallest data regimes in the Gulf are the two a buyer can actually work from. Oman's
Regulatory Framework for National Data Governance and Management is the best-documented GCC regime on readable primary sources: three published documents, thirteen numbered policies, named roles and an explicit three-year sequencing of obligations by priority. Kuwait's CITRA Data Classification Policy v2.3 is shorter but more far-reaching in one respect that no other Gulf instrument matches: it binds private sector entities directly, and it requires them to build and surrender a data catalogue. Both regimes, however, share the regional defect. Oman has not published the calendar-year mapping of its three priority tiers, and Kuwait has not published the execution date from which its two deadlines run. The obligations are legible; the dates are not.
The Omani framework was issued by MTCIT under Ministerial Circular 113/2025, with the policies document dated 1 January 2025. It comprises three documents: the Policies themselves, a Guideline for Establishing a Data Governance and Management Office, and a Guide for Assessing Compliance. That triad is unusual in the Gulf, because it supplies not only the duties but the organisational design to discharge them and the method by which compliance will be judged. Thirteen policies are covered, including Data Catalogue and metadata, Data Classification, Data Quality, Reference and Master Data, and Open Data. One of the thirteen has no counterpart elsewhere in the region: Data Monetisation. Its presence signals that Oman is treating government data as an asset with potential revenue attached rather than solely as a liability to be controlled.
Organisationally, a government unit must establish a Data Governance and Management Office inside its digital transformation division, and must fill four named roles: business data specialist, data technician, data architect and personal data protection officer. The framework permits these to be staffed from existing posts, which is a realistic concession and should be read as such: it lowers the entry cost considerably, and it also means that a unit can report the office as established without having added a single head. Scope covers all government units, with the private sector reached indirectly through sector regulators. A private operator in a regulated Omani sector should therefore expect the framework to arrive through its regulator rather than through the ministry.
Assessment is an annual periodic exercise conducted by the ministry's compliance team under the published Compliance Assessment Guide. The sequencing is the feature most worth copying. The provisions of the thirteen policies are split across three years by priority: Priority 1 provisions in year one, Priority 2 in year two, Priority 3 in year three. That is a genuinely proportionate design and the only staged implementation path published in the Gulf. The defect is that the calendar-year mapping has not been published, so a buyer cannot establish from the instrument which calendar year constitutes year one. Classification runs on four levels borrowed from the Document and Archives Law, Top Secret, Secret, Restricted and Confidential, with Unclassified alongside. The data owner classifies, and the recipient may not alter the classification. No penalty is published.
The CITRA Data Classification Policy v2.3 carries neither an issue date nor a commencement date on its face, and that omission is the central problem with an otherwise clear document. Its scope is what distinguishes it: it binds public and private sector entities, making it the only GCC data-classification instrument that reaches private business directly. Four tiers apply: Public, Private Insensitive, Private Sensitive and Highly Sensitive. Tiers 3 and 4 must be encrypted in transit. An entity must form a data classification team headed by senior management and including the Directors of Information Security and Information Technology, build a classification road map and conduct periodic internal review.
The obligation with the longest tail is clause 3.1.3, under which the data owner must create and maintain a data catalogue including metadata in a unified format, updated periodically. Reporting then makes the catalogue an external deliverable rather than an internal record. The entity focal point submits an implementation report to CAIT every quarter, and under clause 3.4.2.3 CITRA draws periodic compliance reports which must contain the full data catalogue with metadata, the approved classification tiers with their criteria, storage locations by tier and progress against the road map. A buyer should read that as a requirement to keep the catalogue in a state fit to be handed over four times a year, which is a significantly different engineering problem from keeping one for internal use.
Two deadlines exist in the policy and neither can be calculated. Clause 3.4.2.4 requires an implementation
plan within three months of the policy's execution date, and requires entities to be reconciled within two years. The execution date is not published. Both deadlines are therefore unplaceable on a calendar, while the quarterly reporting duty runs regardless. The prudent reading for a private entity in scope is that the two-year reconciliation window has in all likelihood already begun, and that there is no published basis on which to argue otherwise.
One secondary source states that Decision 26 of 2024 repealed the Data Classification Policy. CITRA continues to publish it. The point is unresolved, and the policy is treated here as live because the regulator still publishes it.
Read side by side, Oman and Kuwait make the case that legibility and regulatory weight are independent variables. Neither state has the apparatus of Saudi Arabia or the specificity of Abu Dhabi, and both have produced instruments from which a competent team could write a work plan in an afternoon. They also demonstrate the limits of that achievement. Oman has published a three-year priority sequence without saying when year one began. Kuwait has published two deadlines without publishing the date they run from. Neither publishes a penalty.
The takeaway is twofold. For a buyer in either state, the undated character of the obligations argues for treating them as live now rather than waiting for a date to be announced, because an undated mandatory duty carries no grace period. And for a buyer elsewhere in the Gulf, the Omani Compliance Assessment Guide is the most useful free artefact in the region for structuring a self-assessment, whatever the jurisdiction of the entity using it.
The MIT study does not find that 95 per cent of artificial intelligence projects fail, and the Gartner figure does not say that 60 per cent of them will be abandoned. Both carry qualifiers that are routinely dropped, and a buyer who repeats either without them will lose the argument.
Two numbers now carry most of the weight in Gulf boardroom arguments about data foundations, and both are being quoted incorrectly. The first is the
MIT figure, usually rendered as the finding that 95 per cent of artificial intelligence projects fail. The study did not measure project failure and did not find that. The second is the Gartner figure, usually rendered as a prediction that 60 per cent of artificial intelligence projects will be abandoned. The prediction carries a qualifier, unsupported by AI-ready data, and without it the figure is simply wrong. The value of this chapter is defensive. A buyer who brings either number to a board in its popular form will be corrected by the first person in the room who has read the source, and the correction will cost more credibility than the number was ever worth.
| Figure as usually quoted | What the study actually measured | Sample | The correctly stated finding |
|---|---|---|---|
| 95 per cent of AI projects fail (MIT) | Whether deployed generative artificial intelligence produced sustained productivity or profit-and-loss impact noticed by users or executives at around six months | 153 survey responses, 52 executive interviews and more than 300 publicly disclosed implementations | Only about 5 per cent of custom generative artificial intelligence tools reach production with measurable impact |
| 60 per cent of AI projects will be abandoned (Gartner) | Analyst prediction, with a stated condition on the projects in scope | 1,203 data management leaders, surveyed July 2024 | Through 2026, organisations will abandon 60 per cent of AI projects unsupported by AI-ready data |
| 63 per cent lack the data practices for AI (Gartner) | Whether organisations have the right data management practices for artificial intelligence | 1,203 data management leaders, surveyed July 2024 | 63 per cent either do not have, or are unsure whether they have, the right data management practices for artificial intelligence |
| More than 80 per cent of AI projects fail (RAND) | Root causes of failure, drawn from practitioner interviews | 65 data scientists and engineers, each with five or more years of experience | An outside estimate that RAND cites, not a rate RAND measured |
| Two-thirds say data is the obstacle (McKinsey) | Identification of the primary obstacle to enabling artificial intelligence | Sample reported variously in secondary coverage; the primary document was not read | More than two-thirds of high-performing companies identify data as the primary obstacle to enabling artificial intelligence |
The primary source is The GenAI Divide: State of AI in Business 2025, published by MIT Project NANDA in July 2025, authored by Challapally, Pease, Raskar and Chari, running to 26 pages. The document describes itself as preliminary findings and it is not peer reviewed. Its evidence base combines 153 survey
responses, 52 executive interviews and more than 300 publicly disclosed implementations. What it set out to establish was whether generative artificial intelligence, once deployed, produced sustained productivity or profit-and-loss impact that users or executives actually noticed at around the six-month mark. That is a question about realised benefit, not about delivery failure. A project that was delivered on time and on budget, and that nobody could show had moved a number six months later, falls on the wrong side of the MIT line while being a successful project by any conventional measure.
Two further limits matter for a Gulf reader specifically. The study contains no geographic breakdown and no public-sector analysis, so it says nothing whatever about the Gulf, and nothing about government, which is the largest buyer of data and analytics capability in every one of the six states. Critics have also noted that the 95 per cent derives from a subset of the evidence base rather than from the whole of it. The defensible version of the finding is the one the study itself frames: only about 5 per cent of custom generative artificial intelligence tools reach production with measurable impact. That is a serious claim and it supports the data foundations argument perfectly well. It simply is not the claim being repeated.
The Gartner source is a press release of 26 February 2025 titled Lack of AI-Ready Data Puts AI Projects at Risk, quoting Roxane Edjlali. The prediction is that through 2026, organisations will abandon 60 per cent of artificial intelligence projects unsupported by AI-ready data. The final five words are the whole of it. The prediction is conditional on a population of projects that lack AI-ready data, and it says nothing about the abandonment rate of projects that have it. Dropping the qualifier converts a conditional statement about a subset into a universal claim about the whole, which is both wrong and, awkwardly for anyone making the foundations argument, self-defeating: the conditional version is the one that actually attributes failure to data.
The original release carrying the 30 per cent prediction was not re-verified. It should be cited as a Gartner prediction without a date attached unless the buyer locates the release.
The companion statistic from the same survey of 1,203 data management leaders, conducted in July 2024, is cleaner and more useful. Sixty-three per cent of organisations either do not have, or are unsure whether they have, the right data management practices for artificial intelligence. The two clauses should be kept together. The uncertainty is itself the finding, because an organisation that cannot say whether its data practices are adequate has already answered the question. Gartner separately predicts that through 2026, 30 per cent of generative artificial intelligence projects will be abandoned after proof of concept, owing to poor data quality, inadequate risk controls, escalating costs or unclear business value. That prediction has four stated causes and data quality is one of four.
RAND's The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed, published 13 August 2024 by Ryseff, De Bruhl and Newberry, rests on interviews with 65 data scientists and engineers each carrying five or more years of experience. It identifies five root causes, of which the second is that the organisation lacks the data needed to train an effective model. That is a well-sourced qualitative finding and a buyer can use it. What a buyer cannot use is the figure most often attached to the paper, that more than 80 per cent of artificial intelligence projects fail. RAND cites that as an outside estimate. It did not measure it, and it should not be attributed to RAND. McKinsey's position that more than two-thirds of high-performing companies identify data as the primary obstacle to enabling artificial intelligence is quotable as stated. The adoption figures circulating from its State of AI work, 88 per cent adoption against 7 per cent fully scaled, come from secondary coverage with the sample reported variously as 1,933, 1,993 and about 1,500. The primary document was not read, so no sample figure should be cited with them.
The takeaway is a rule rather than a number. Any figure brought into a Gulf data argument should be stated with its qualifier, its sample and its date, and should be dropped entirely if any of the three cannot be produced. The evidence that data foundations determine artificial intelligence outcomes is strong enough to survive that discipline. The headline versions are not, and they are the ones a sceptical finance director will check first.
One Gulf state publishes an aggregate data maturity score and none publishes entity-level scores. On the one substantial regional survey, artificial intelligence spending is rising at more than a hundred per cent a year while fewer than one organisation in seven has replaced the systems its data sits in.
The honest answer to where the Gulf stands on data foundations is that almost nobody will say. No Gulf state publishes entity-level data maturity scores. The United Arab Emirates is the only one of the six to publish even an aggregate, and Saudi Arabia has operated a purpose-built national index since November 2023 without releasing a result from it. What can be assembled instead is a composite: one published aggregate, one set of open dataset counts, one vendor-commissioned regional survey and a talent picture with a single sourceable figure in it. Taken together they describe a region spending very fast on artificial intelligence and moving slowly on the foundations beneath it, which is the gap this edition is named for.
The UAE Data Maturity Index is operated by the Federal Competitiveness and Statistics Centre and was presented to the United Nations Statistical Commission Network on Economic Statistics on 4 June 2024. It reported an average data maturity of federal entities of 80 per cent. The methodology is published in full, which is rarer and more valuable than the score. Four bands apply: Basic from 0 to 49, Developing from 50 to 69, Mature from 70 to 95 and Pioneer from 96 to 100. Three weighted pillars produce the score: Data Regulatory and Operational Framework at 10 per cent, Data Quality Standards at 40 per cent and Data Availability and Exchange at 50 per cent. The index draws on 233 million data records received monthly from federal entities. Entity ranks from 1 to 15 are shown, but the names and individual scores are not public.
The weighting repays attention, because it tells a buyer what the federal government considers data maturity to consist of. Half the score sits in availability and exchange, two-fifths in quality standards and only a tenth in the regulatory and operational framework. An entity that has written policies and appointed officers, and has not made its data available or demonstrably accurate, scores badly on this index by design. That is a useful corrective to a market that often measures governance documents instead.
Against that single disclosure sits an extensive silence. No Saudi National Data Index result has been found for any year; the only entity-level outcome located anywhere was the recognition of the Ministry of Health as an advanced entity on 26 December 2024, without a score. The Digital Dubai Data Maturity Framework was launched at GITEX 2022 with four pillars, Data Governance, Data Sharing, Data Analysis and Decision-Making, and Data Quality, and its scores are not public; a data quality initiative began with eleven government entities in July 2024. Qatar Data Trust Index scores are not public. No count of chief data officers appointed exists for any GCC state. Dubai approved the appointment of 22 Chief Artificial Intelligence Officers on 9 June 2024, and those are artificial intelligence officers, not chief data officers: the two should not be conflated, and the appointment of the former is frequently offered as evidence of the latter.
| State | Open datasets, as at 6 May 2025 | Later figure where published |
|---|---|---|
| Saudi Arabia | 10,916 | Not published |
| United Arab Emirates (Bayanat) | 3,146 | Not published |
| Bahrain | 448 | 515 datasets from 42 publishers, currently shown on data.gov.bh |
| Qatar | 159 | More than 1,100, on relaunch, 27 June 2025 |
| Oman | 60 | Not published |
| Kuwait | No centralised portal | Not published |
| Gulf total | 14,729 | Superseded for two states |
Counts taken from the national portals on a single date, 6 May 2025, and reported by ORF Middle East on 20 May 2025. The comparison is already superseded for Qatar and Bahrain, as shown. Dataset counts measure publication, not quality or maturity.
The spread is the point. Open data is an imperfect proxy for maturity, since a count says nothing about whether the datasets are current, documented or used, but it is one of the very few measures in the region that is both public and comparable, and the distribution broadly tracks the other indicators. On the United Nations E-Government Survey 2024, published in the UN DESA Technical Appendix of 30 October 2024, Saudi Arabia ranked 6th with an index score of 0.9602, the UAE 11th at 0.9533, Bahrain 18th at 0.9196, Oman 41st at 0.8576, Qatar 53rd at 0.8244 and Kuwait 66th at 0.7812. On the Oxford Insights Government AI Readiness Index 2025, Saudi Arabia placed 15th globally at 69.79, the UAE 23rd at 66.87 and Qatar 54th; the
data-dimension sub-scores are not public, so neither index can be decomposed to isolate data capability.
The most directly relevant Gulf evidence comes from the ServiceNow Enterprise AI Maturity Index 2026, built on ThoughtLab fieldwork covering 4,500 executives across 19 countries, including 100 in Saudi Arabia and 100 in the United Arab Emirates, and reported by Middle East AI News on 18 August 2026. Seventy-seven per cent of UAE executives and 67 per cent of Saudi executives named inadequate data accuracy, access and management as a major barrier to scaling artificial intelligence. Maturity scores were 48 out of 100 for the UAE and 50 out of 100 for Saudi Arabia. Artificial intelligence spending rose 105 per cent year on year in the UAE and 124 per cent in Saudi Arabia. And only 14 per cent of UAE organisations and 13 per cent of Saudi organisations had replaced legacy systems with platforms capable of integrating artificial intelligence.
The Gulf samples are 100 executives per country and the study is vendor-commissioned. The direction is credible and the precision of the percentages should not be relied upon. A Cloudera survey of Saudi data readiness, with its sample size not published, found 62 per cent of Saudi information technology leaders citing data access restrictions as a major barrier and 32 per cent reporting complete visibility into their data.
On people the public record is thinner still. The only Gulf figure on specialist capability is Saudi: SDAIA has trained more than 779,000 citizens, including 9,775 specialists and 260 scientists and experts, per the 2024 Vision 2030 report (SPA N2311819, 5 May 2025), against an announced NSDAI target of approximately 20,000 data and artificial intelligence specialists by 2030. Later counts of 11,000, 14,000 and 14,495 circulate widely and trace to no dated primary release, and are not used here. No national labour office in Saudi Arabia or the UAE publishes vacancy counts at occupation level, so the number of open data engineering roles in the Gulf is not public and cannot be estimated from official sources. GulfTalent reports artificial intelligence skills appearing in about 3.4 per cent of professional vacancies across the UAE, Saudi Arabia and Qatar in the first half of 2026, up from 1.2 per cent in 2022, drawn from 118,000 direct-employer vacancies
on its own job board; that measures artificial intelligence skills, not data engineering. Korn Ferry's Q1 2026 survey of 105 chief human resources officers across the six states found 93 per cent exploring, piloting or deploying artificial intelligence and 1 per cent considering themselves fully ready to scale, a finding that is human-resources-scoped rather than data engineering.
The takeaway is that a Gulf buyer cannot benchmark against peers, because peer scores are not published anywhere in the region. The practical substitute is to self-score against the UAE index, whose bands and weights are public, and to treat the 80 per cent federal average as the only published reference point in the Gulf. It is a UAE federal figure and should be used as such, not as a regional norm.
A Snowflake Standard credit costs 35 per cent more in Dammam than in Northern Virginia and 30 per cent more in Azure UAE North. On the published evidence the premium is a Snowflake characteristic rather than a regional one, because most platform vendors do not publish Gulf rates at all.
There is a Gulf cloud data premium, it is documented in one vendor's published rate card, and it is larger than most budgets assume. On the Snowflake Service Consumption Table effective 9 November 2026, a Standard on-demand credit costs 35 per cent more in GCP Middle East Central 2 at Dammam than in AWS US East at Northern Virginia, and 30 per cent more in Azure UAE North. Whether that premium generalises cannot be established, and the reason is the second finding of this chapter: four of the five vendors examined render their price tables in client-side JavaScript, so several Middle East rates are not published in retrievable form at all. Microsoft Fabric publishes no capacity price in any region, BigQuery publishes no Gulf rate, and Amazon Web Services publishes no Middle East rate for Redshift, Glue, EMR or Lake Formation.
Two features of the table deserve more attention than the headline premium. The premium compounds up the edition ladder: a buyer on Business Critical in Dammam pays USD 6.50 against USD 4.00 in Northern Virginia, so the hosting decision and the edition decision multiply against each other. And Dammam storage is flat at USD 30.00 per terabyte per month across all seven capacity tiers, where the usual expectation is that volume attracts a lower unit rate. A business case that assumed storage economies of
scale at the point of commitment will not find them in that region.
| Region | Standard | Enterprise | Business Critical | VPS | Storage per TB per month |
|---|---|---|---|---|---|
| AWS US East, Northern Virginia | USD 2.00 | USD 3.00 | USD 4.00 | USD 6.00 | USD 23.00 |
| Azure UAE North | USD 2.70 | USD 4.00 | USD 5.40 | USD 8.10 | USD 25.40 |
| GCP Middle East Central 2, Dammam | USD 3.25 | USD 4.90 | USD 6.50 | USD 9.75 | USD 30.00 |
Snowflake on-demand credit prices and storage rates from the Service Consumption Table effective 9 November 2026, retrieved 10 November 2026. No rate shown was derived from any other rate. Dammam storage is flat at USD 30.00 across all seven capacity tiers with no tier discount. AWS me-central-1 in the UAE and me-south-1 in Bahrain are not listed at all. Capacity credit prices are not public as rates.
| Azure Databricks Premium, per DBU-hour | UAE North | Qatar Central |
|---|---|---|
| Jobs Compute | USD 0.30 | USD 0.30 |
| All-purpose | USD 0.55 | USD 0.55 |
| SQL Compute Pro | USD 0.74 | USD 0.74 |
| Serverless SQL | USD 0.95 | Not published |
| Delta Live Tables, Core | USD 0.30 | USD 0.30 |
| Delta Live Tables, Pro | USD 0.38 | USD 0.38 |
| Delta Live Tables, Advanced | USD 0.54 | USD 0.54 |
Databricks presents the opposite picture, and it is the evidence against treating the Snowflake premium as a regional fact. The rates are identical in both Gulf regions and carry no regional uplift at all. Qatar Central publishes no serverless rate, which is a capability difference rather than a pricing one. The conclusion a buyer should draw is that the Gulf premium is a property of a particular vendor's regional rate card rather than of Gulf infrastructure, and that it must therefore be tested vendor by vendor rather than assumed.
Rates taken from the Microsoft retail prices application programming interface that bills the service, retrieved 10 November 2026. Databricks is not available in any AWS Middle East region, and GCP me-central2 at Dammam is supported, per documentation updated 21 October 2026. Databricks own-brand DBU list prices for AWS and GCP, Dammam included, are not public by static fetch.
Google BigQuery publishes a complete rate card for us-central1: on-demand at USD 6.25 per tebibyte scanned with the first tebibyte each month free,
Standard edition at USD 0.04 per slot-hour, Enterprise at USD 0.06 and Enterprise Plus at USD 0.10, with one-year and three-year commitment discounts, and active logical storage at USD 0.000031507 per gibibyte-hour. The Gulf regions me-central1 at Doha, me-central2 at Dammam and me-west1 all appear in the region selectors, so rates exist, but the served page returns us-central1 only. Gulf BigQuery rates are therefore not public by static fetch. Two figures are published and region-blind, and they are the ones a Gulf architect should price carefully: cross-region transfer into the Middle East at USD 0.11 per gibibyte from North America, Europe or Asia, and USD 0.08 per gibibyte for Middle East to Middle East.
Microsoft Fabric publishes its full SKU ladder from F2 to F8192 with capacity units attached, and every price cell renders as a dash. Pay-as-you-go and reserved F-SKU prices are not public by static fetch in any region, not merely in the Gulf, while the region selector lists UAE North, UAE Central and Qatar Central: capacity is available in three Gulf regions at prices that cannot be read. One reservation line is retrievable for UAE North, where the meter Fabric Capacity CU shows a one-year retail price of 1146.0 and a three-year price of 3438.0, with the unit of measure labelled by the interface as 1 Hour. OneLake storage in UAE North is published per gigabyte per month at USD 0.0253 for Hot, USD 0.015 for Cool, USD 0.005 for Cold and USD 0.264 for Cache.
| AWS service, US East Northern Virginia | Published rate |
|---|---|
| Glue ETL Spark | USD 0.44 per DPU-hour |
| Glue Flex | USD 0.29 per DPU-hour |
| Glue crawlers | USD 0.44 per DPU-hour |
| Glue Data Quality, recommendation and evaluation | USD 0.44 per DPU-hour, two-DPU minimum |
| Glue Data Catalog, metadata objects | First 1,000,000 free, then USD 1.00 per 100,000 per month |
| Glue Data Catalog, requests | First 1,000,000 per month free, then USD 1.00 per million |
| Lake Formation permissions, tag-based access control, cross-account sharing | No charge |
| Redshift Serverless | USD 0.375 per RPU-hour |
| Redshift ra3.4xlarge | USD 3.26 per node-hour, from AWS's own worked example |
| Redshift managed storage | USD 0.024 per GB-month |
| Redshift Spectrum | USD 5.00 per TB scanned |
AWS pricing pages updated 5 November 2026. The RA3 node tables render empty. Middle East rates in me-central-1 and me-south-1 for Redshift, Glue, EMR and Lake Formation are not public by static fetch, and the only sourceable AWS statement on the point is that pricing varies by Region. The rates above are therefore a Virginia baseline and not a Gulf quotation.
Below the rate cards the position worsens. No published register of GCC data contract awards exists, and not one award across the six states could be sourced with buyer, scope and value published together by the buyer. Etimad is blocked to automated access by CAPTCHA and Qatar's Monaqasat returns an empty body. The one Etimad aggregate available is that contracts awarded through the platform exceeded SAR 100bn in 2023, with no category breakdown, and the platform excludes Aramco, SABIC, STC and PIF portfolio companies. The only Gulf contracts carrying a named government buyer and a published value are Tadawul disclosures by Al Moammar Information Systems with SDAIA as buyer, set out below, and every one of them is information technology infrastructure supply rather than data governance or analytics work.
| Disclosed value | Scope, SDAIA as buyer | Term or date |
|---|---|---|
| SAR 58.7m including VAT | SDAIA artificial intelligence network enhancement | 36 months |
| SAR 64.63m | National Information Center infrastructure support | Not stated |
| SAR 83.95m | Scope not stated in the sourced disclosure | Signed 20 November |
| SAR 98.04m | Scope not stated in the sourced disclosure | Signed 19 November |
| SAR 134.84m | Framework agreement | 36 months, awarded 17 March 2026 |
| SAR 228m | Expansion of the Naqaa data centre, Riyadh | Not stated |
Signing dates do not reconcile across reports, which is why two rows above carry a month without a year. One further contract is public without a value: Elm's subsidiary Thiqah holds an agreement with the Ministry of Commerce to digitise ministry services on a revenue-sharing basis, with no fixed total and a five-year term from 3 September 2026.
Three categories should be kept separate and routinely are not. Spent: nothing, because no Gulf government publishes realised spend on data or analytics programmes. Committed with a disclosed value: only the Al Moammar contracts above. Announced only: Snowflake's USD 300m in Saudi Arabia through 2030, stated on 3 November 2026, Databricks at USD 300m over five years, Huawei Cloud at USD 400m, Alibaba Cloud at USD 500m and Abu Dhabi's AED 13bn digital strategy. The Abu Dhabi Department of Government Enablement agreement with Microsoft and Core42 on sovereign cloud has no disclosed value, and the AED 13bn figure attached to it in
secondary coverage is the whole-of-government Digital Strategy 2025 to 2027 plan rather than the contract. That is a common and consequential error. On analyst framing, Gartner put MENA information technology spending at USD 169bn in 2026, up 8.9 per cent, with data centre systems the fastest-growing segment at 37.3 per cent to about USD 13bn and software up 13.9 per cent to USD 20.4bn (press release, 4 August 2025); MENA there includes Egypt, Morocco and Algeria. IDC put MEA information technology spending at USD 155bn in 2025 with 5 per cent forecast growth in 2026. No Gartner or IDC data-and-analytics-specific Middle East figure is public.
| Role, United Arab Emirates | AED per month, basic plus fixed allowances |
|---|---|
| Data Governance Director | 74,000 to 85,000 |
| Data Architect | 48,000 to 59,000 |
| Data Engineer, industry | 27,000 to 44,000 |
| Data Engineer, start-ups | 25,000 to 35,000 |
| Data Scientist | 44,000 to 55,000 |
Cooper Fitch UAE Salary Guide 2025, published December 2024. These are recruiter estimates, not a statistical survey. The Robert Walters Middle East Salary Survey 2026 contains no Data Engineer and no Data Architect line, and its Saudi technology table is identical line for line to its UAE table under an AED heading, so no 2026 Saudi figure should be taken from it. Hays, Michael Page, Cooper Fitch KSA and Mercer are gated, and role-level figures for Qatar, Oman, Bahrain and Kuwait are not public from any named guide.
Four widely quoted market figures should not appear in a buyer's business case at all. PwC's USD 320bn is a GDP-contribution model from a press release of 11 February 2018 covering the Middle East rather than the GCC. Statista's MEA big data item is a republished third-party forecast released in March 2017 with paywalled values. Six commercial houses size the same 2025 Middle East data analytics market between USD 4bn and USD 17.75bn, a spread of 4.4 times. And Presight's AED 2.48bn second-quarter 2026 order intake and AED 4.9bn backlog are neither contract values nor awards.
The takeaway is that Gulf data platform budgeting cannot be done from public list prices, because for three of the five vendors there are none to read in the region. What can be done is the arithmetic that the Snowflake table permits: price the same workload in a Gulf region and in Northern Virginia, carry the difference as an explicit sovereignty cost, and put it in front of whoever decided that the data must stay in the country. That decision is usually correct. It is rarely costed.
The strongest freely available data maturity model for a Gulf buyer is the UAE Federal Data Maturity Index, which publishes its bands and its pillar weights in full. The better-known alternatives are licensed, unreachable or retired.
A buyer wanting to score its own data maturity has six candidate models and should choose the UAE Federal Data Maturity Index. It is free, it is Gulf-native,
and alone among the options it publishes both its scoring bands and the weights of its pillars, which means a self-assessment produced against it can be reproduced and challenged. The alternatives fail on availability rather than on quality: the best-known international model, DCAM, is available only to member firms of the body that owns it; the most widely cited name, CMMI's data management model, appears to have been retired; and Gartner's own model is not sourced in the open at all. The recommendation that follows is therefore not that the UAE index is the most sophisticated instrument available. It is that it is the most sophisticated instrument a Gulf buyer can actually obtain and apply.
| Model | Structure | Availability | Recommended use |
|---|---|---|---|
| UAE Federal Data Maturity Index | Four bands, Basic 0 to 49, Developing 50 to 69, Mature 70 to 95, Pioneer 96 to 100; three weighted pillars, Regulatory and Operational Framework 10 per cent, Data Quality Standards 40 per cent, Data Availability and Exchange 50 per cent | Free; bands and weights fully published | The scoring spine. Self-score against it at no cost |
| Saudi National Data Index maturity scale | Six levels, Level 0 Absence of Capabilities to Level 5 Pioneer; 42 questions across 14 domains | Free; methodology published, but the compliance component depends on NDMO specifications whose operative text is not reachable | Useful as a question set for a Saudi entity. Not usable as a complete assessment |
| Oman Guide for Assessing Compliance | Thirteen policies with provisions tiered Priority 1 to Priority 3 across three years | Free; published | The control checklist and the staging plan, usable in any Gulf jurisdiction |
| DAMA DMBOK, 2nd edition 2017 | A body of knowledge. Its Data Management Maturity Assessment chapter sets out a six-level scale from Level 0 absence through Initial or Ad Hoc, Repeatable, Defined and Managed to Optimised, assessed on activity, tools, standards, and people and resources | The book is purchasable; the framework is not licensed for assessment use | The vocabulary and the domain taxonomy. Not a scoring model |
| Model | Structure | Availability | Recommended use |
|---|---|---|---|
| EDM Council DCAM | Version 3 has 8 components, 34 capabilities and 101 sub-capabilities, each rated on a six-level maturity scale; v3.1 current as at April 2026 | Licensed. Available only to EDM Council member firms through the Knowledge Portal. Training and certification listed at USD 1,370 per seat for members, with a USD 950 upgrade course. Membership fees not public | Appropriate only where the buyer is already a member firm |
| CMMI Data Management Maturity | Released August 2014. Five levels and 25 process areas across six capability areas: data governance, data quality, data operations, data architecture, data development and data security | ISACA is reported to have discontinued support in January 2022 and the model page removed. Treated as retired, not officially confirmed | Not recommended |
| Qatar Data Trust Index | Seven dimensions and 50 sub-indicators, producing a Data Quality Score and a Data Compliance Score, with a stated action threshold at 85 | Methodology document not reachable | Binding on Qatari government entities. Not usable as a self-assessment by anyone else |
The case for the UAE index rests on three properties. It is transparent: the four bands and the three pillar weights are published, so a score can be recomputed by anyone who disputes it. It is opinionated in a way that is useful: half the weight sits in Data Availability and Exchange and two-fifths in Data Quality Standards, leaving only a tenth for the regulatory and operational framework, which prevents an organisation scoring well on the strength of policy documents alone. And it is Gulf-native, built by a Gulf statistical authority for Gulf government entities, which makes it easier to defend internally than an American or European framework. Its limitation is one it shares with every other instrument in the region: entity names and individual scores are not published, so the index yields a score and no peer group. A buyer can establish that it sits in the Developing band. It cannot establish whether that is better or worse than its neighbours, and the published federal average of 80 per cent is the only external reference point available anywhere in the Gulf against which to read its own result.
DCAM, the Data Management Capability Assessment Model of the EDM Council, is the most granular instrument on the list. Version 3 sets out 8 components, 34 capabilities and 101 sub-capabilities, each rated on a six-level maturity scale, and version 3.1 is current as at April 2026. That granularity is genuine and it comes with a gate: DCAM is available only to member firms through the Council's Knowledge Portal. Training and certification are listed at USD 1,370 per seat for members, with a USD 950 upgrade course, and membership fees are not published. For an institution that is already a member, typically a bank, DCAM is the right choice and this chapter's recommendation does not apply. For everybody else the position is that the assessment cannot be started without a procurement decision, and a Gulf buyer whose immediate problem is to establish a baseline will reach the baseline faster with the UAE index.
DAMA DMBOK, second edition 2017, is the most widely held reference in the field and the most frequently misused. It is a body of knowledge, not a scoring model. Its Data Management Maturity Assessment chapter does set out a six-level scale, Level 0 for absence of capability through Initial or Ad Hoc, Repeatable, Defined and Managed to Optimised, assessed across activity, tools, standards, and people and resources. Sources disagree over whether the level names include Measured, which is itself a reason not to present the scale as authoritative. The book is purchasable and the framework is not licensed for assessment use. The right role for DMBOK in a Gulf programme is as the vocabulary and the domain taxonomy: it supplies the language in which the Abu Dhabi, Omani and Saudi domain lists can be reconciled with one another, which is a genuine service and not a maturity score.
The CMMI Data Management Maturity model, released in August 2014 with five levels and 25 process areas across six capability areas, remains the most quoted name in Gulf tender documents. ISACA is
reported to have discontinued support for it in January 2022, and the model page has been removed. That report is not officially confirmed, which is the proper caveat, but the practical conclusion does not change: a model whose owner appears to have withdrawn it and whose documentation is no longer published should not be written into a specification. A buyer who finds it named in a draft requirement should substitute the UAE index and record why.
The takeaway is a three-part construction that costs nothing. Use the UAE Federal Data Maturity Index as the scoring spine, because its bands and weights are public. Use Oman's Guide for Assessing Compliance as the control checklist and the staging plan, because it is the only published instrument in the Gulf that sequences obligations by priority across three years. Use DAMA DMBOK for the vocabulary. Then overlay the binding instrument of the buyer's own jurisdiction, which in Abu Dhabi, Oman and Kuwait is readable and in Saudi Arabia and Qatar is not. Nothing in that stack requires a licence, and the whole of it can be assembled from documents a buyer can download today.
Almost no GCC data instrument carries a published deadline. Across the six states the only dated obligations in the 2026 to 2028 window are recurring reporting cycles with no fixed day, and the hardest dated data obligation facing a Gulf group with Indian operations is Indian.
The most striking finding in this edition is an absence. Across the six GCC states, almost no data instrument carries a published deadline. The obligations that bind now are recurring reporting cycles, and of those only one, the Saudi third-quarter audit, even names a quarter. Oman tiers its provisions across three years without publishing which calendar year is year one. Kuwait sets two deadlines that run from an execution date it has not published. Abu Dhabi requires regular reporting and defines regular nowhere. Qatar's review cycle is undated. There is no dated Saudi NDMO deadline, no dated Qatar National Data Policy deadline and no dated Bahraini obligation of any kind. The hardest dated data obligation in the window belongs to a jurisdiction outside the Gulf, and falls in May 2027 in India.
| When | Obligation | Who is bound | Status |
|---|---|---|---|
| Q3 of 2026, 2027 and 2028 | NDMO compliance audit and Entity Compliance Report | All Saudi public entities | Recurring and binding. No fixed day published |
| Annually | MTCIT periodic compliance assessment under the Compliance Assessment Guide | All Omani government units | Recurring and binding. Priority 1 provisions in year one, Priority 2 in year two, Priority 3 in year three. Calendar-year mapping not published |
| Every quarter | Implementation report to CAIT, including the full data catalogue with metadata | Kuwaiti public and private sector entities in scope | Recurring and binding. No fixed day published |
| Frequency stated only as regular | Programme status report against Data Management Programme Plan milestones | Abu Dhabi government entities and their contractors | Recurring and binding. No interval published |
| Cycle not dated | Internal review against the National Data Standards and Qatar Data Trust Index scoring | Qatari government entities | Recurring. Cycle not published |
| During 2026 or 2027, no published date | PDPL Executive Regulations expected from the Federal Authority for Artificial Intelligence and Data | UAE controllers and processors | A forecast, not an obligation |
| Two years from the policy's execution date | Entities reconciled with the CITRA Data Classification Policy | Kuwaiti public and private sector entities in scope | Not placeable, because the execution date is not published |
| About 13 November 2026 | DPDP Rules 2025, Rule 4, consent manager registration | Consent managers in India | Dated and notified |
| About 13 May 2027 | DPDP Rules 2025, Rules 3, 5 to 16, 22 and 23, and the substantive sections | Indian data fiduciaries, including Gulf groups with Indian operations | Dated and notified. The hardest dated data obligation across the jurisdictions examined |
It would be comfortable to read the absence of deadlines as a sign that Gulf data obligations are soft. The opposite is the better reading. A deadline is a concession. It tells the regulated party when the duty begins and, by implication, that it does not bind before. An instrument that is mandatory on issue and sets no deadline has been in force since the day it was issued, which means there is no grace period to rely on and no transitional defence to run. Oman's Circular 113/2025 has bound all government units since January 2025. Qatar's classification policy has bound government and vital-sector entities since 7 May 2023 and its National Data Policy since October 2024. Dubai's Resolution 2 has bound since 2017. Abu Dhabi's policy
is mandatory on its own terms and its issue date is not even published. In each case the entity is already late if it is not compliant, and nothing in the instrument says how late.
Two items in that list are privacy instruments rather than data management, and the distinction holds throughout this edition: the DIFC and ADGM regimes govern personal data and impose no catalogue, quality or lineage duty. The UAE federal picture is the one genuinely awaiting an event. Federal Decree-Law 45 of 2021 has been in force since 2 January 2022 and its Executive Regulations had still not been issued as at early October 2026. A claim of a 1 January 2027 compliance deadline circulates on vendor sites, cites no instrument and is contradicted by the majority of sources: it should not be used. The Federal Authority for Artificial Intelligence and Data, created by Cabinet decision on 14 June 2026 and absorbing the Data Office, is the body expected to issue them, and the expectation is a forecast rather than an obligation.
India's data-management instrument, by contrast, behaves exactly like the Gulf's. The National Framework
for Data Sharing was approved by the Committee of Secretaries on 15 May 2026, having been prepared by MeitY, and rests on four pillars. It requires every Ministry and Department to constitute a Data Governance Committee, with State committees chaired by Chief Secretaries. It carries no commencement date, no deadline and no penalty: it is administrative rather than statutory. The live tracker shows State committees constituted in about 19 of 36 States and Union Territories, and State data sharing policies published in nine. An earlier draft National Data Governance Framework Policy, from 2022, was never notified at all. The pattern is therefore not a Gulf peculiarity. Across eight jurisdictions, personal data attracts dates and penalties, and data management attracts committees and audits.
For a Gulf group with Indian operations, and many have them, the Indian calendar is firmer than anything in the Gulf. The Digital Personal Data Protection Act 2023 was enacted on 11 August 2023 with commencement deferred, and the DPDP Rules 2025 were notified on 13 November 2025 with a phased schedule. Rule 4, covering consent manager registration, takes effect about 13 November 2026. Rules 3, 5 to 16, 22 and 23, together with the substantive sections, take effect about 13 May 2027. That May 2027 date is the hardest dated data obligation across all the jurisdictions examined here, Gulf and Indian together, and it is the one around which a group programme can actually be sequenced.
The takeaway is to stop planning against deadlines in the Gulf, because there are almost none, and to plan instead against the audit calendar. For a Saudi public entity that means the third quarter of every year. For a Kuwaiti entity, public or private, it means a quarterly submission that must include the catalogue itself. For an Omani government unit it means an annual ministry assessment against a priority tier whose year is not published, which argues for treating all three tiers as live. For an Abu Dhabi entity and its suppliers it means programme reporting at an interval the entity will have to fix for itself. In every case the compliance date is now, and the only question the instrument leaves open is when somebody will ask. For a group with Indian operations the sequencing is simpler, because May 2027 is fixed and can be worked backwards from. The Gulf offers no such anchor, and an organisation waiting for one will be waiting during the period in which it is already in default.
Ten dated, falsifiable judgements on Gulf data engineering in 2027: each written to be graded, hit, miss or partial, in the 2028 Scorecard.
Ten judgements on Gulf data engineering in 2027, each dated, falsifiable and confidence-rated. The reasoning for each follows.
These are reasoned judgements, not guarantees. Each is written to be graded, hit, miss or partial, in the 2028 Scorecard, with a running hit-rate reported by confidence band. High-confidence Calls carry structural or already-visible support; lower-confidence Calls turn on execution or disclosure that cannot yet be observed.
The governance scaffolding exists (SDAIA integrates 360+ systems and Qatar's programme stood up in January 2025), but production-grade cross-entity sharing is an engineering milestone beyond governance, and the talent constraint makes the pace uncertain [SDAIA, 2025; NPC, 2025].
By end-2027, at least three GCC national data platforms (the SDAIA National Data Bank tier, Qatar's National Data Program, and a UAE federal or emirate platform) reach production-grade cross-entity data sharing.
Confidence: Medium · Resolves end-2027
The GCC data-engineering talent gap remains binding through 2027, with demand outstripping locally available data engineers despite localisation programmes.
Confidence: High · Resolves through 2027
This is a structural, long-apprenticeship shortage that capital cannot clear at speed; even Saudi Arabia's own strategy concedes the scale, targeting 20,000+ specialists by 2030. The constraint is human and slow-moving [NSDAI, 2020].
Sovereignty-by-design, classification-driven residency, becomes a default requirement in requests for proposal for GCC government data platforms by end-2027.
Confidence: High · Resolves end-2027
The direction is already set by Abu Dhabi's 100% sovereign-cloud target, vendor offerings (Microsoft Cloud for Sovereignty, Oracle Sovereign Cloud) and the Dubai Data Manual's mandatory standards of July 2026 [DGE, 2025; DDSE, 2026].
The Microsoft Saudi Arabia East region anchors a wave of KSA data-platform migrations through 2027, graded on migration volume rather than on the availability date, which is targeted for November 2026 and unconfirmed.
Confidence: Medium–High · Resolves through 2027
Construction on three sites was complete by December 2024 and customer availability is targeted for November 2026, unconfirmed as at 9 November 2026; a new in-region hyperscaler with sovereign controls is the natural anchor for compliant migrations, though timing risk remains [Microsoft, 2024–26].
A majority of GCC enterprise generative-AI initiatives launched in 2025–26 stall or are re-scoped in 2027 pending data-foundation work.
Confidence: Low–Medium · Resolves through 2027
The base rate is contested: the MIT Project NANDA figure measures organisations recording no profit-and-loss impact rather than pilots failing, and 2026 surveys spread widely on returns. Gartner puts failure near 50%, and the Gulf's foundation lags its compute; the "majority" threshold and the contested base rate are what temper confidence [MIT NANDA, 2025; Gartner, 2026].
At least one GCC national entity publicly reports an AI project failure or rescope explicitly attributed to data quality or engineering by end-2027.
Confidence: Medium · Resolves end-2027
Given the failure base rates and the region's AI ambition, such a failure is probable; the uncertainty is whether it is disclosed publicly and attributed explicitly to data rather than to other causes [Gartner; McKinsey, 2025–26].
System-integrator-led builds begin shifting towards in-house platform teams at at least two flagship GCC entities by 2027.
Confidence: Low–Medium · Resolves through 2027
The platform logic favours in-house ownership and consolidation is visible in infrastructure, but the talent bottleneck directly constrains the ability to staff in-house teams, which caps confidence [DGE, 2025; Palo Alto Networks, 2026].
FinOps-for-data becomes a named function or explicit line item at major GCC entities as cloud-data spend outpaces value in 2027.
Confidence: Medium · Resolves through 2027
The spend-value gap is now large enough to force scrutiny ($44.5bn projected cloud waste, $12.9m per-org data-quality cost), and cheap power has masked inefficiency that finance functions will begin to meter [Harness, 2025; Gartner via IBM, 2026].
A data-mesh or data-product operating model is adopted at at least one GCC anchor (Aramco, ADNOC or a major bank) by end-2027.
Confidence: Medium · Resolves end-2027
The largest Gulf enterprises have the scale, the domain complexity and the in-house engineering depth to adopt a data-product model; the constraint is organisational change rather than capability at these specific anchors.
GCC data-centre live IT capacity passes 3 GW by end-2027, but data-engineering maturity lags the megawatt buildout. The original 1.7 GW bar was passed at publication: Knight Frank's 2026 outlook already places the Middle East at 2.2 GW in 2026.
Confidence: High · Resolves through 2027
Live capacity has run about 1.2 GW at end-2025 to 2.2 GW in 2026 and 3.2 GW forecast for 2027 [Knight Frank, 2026] and is compounding, while the foundation-gap evidence shows maturity trailing; the divergence of the two curves is the edition's thesis in one line [Knight Frank / DC Byte; MIT NANDA, 2025].
Each Call is constructed to be falsifiable, dated to a resolution year, confidence-rated, and graded in the following edition.
Volge Quarterly exists to make a forward-look that can be held to account, and the method is designed to that end. A candidate Call must satisfy four tests before it is admitted: it must be falsifiable, so that a reader can state in advance what evidence would prove it wrong; it must be dated, resolving within a stated year; it must be grounded, tracing to sourced and year-stamped evidence in the preceding parts; and it must be confidence-rated, so that the reader knows how much weight the editors place on it. Calls that fail any test are set aside as watchlist items rather than dressed up as judgements.
Four bands are used. High is reserved for Calls with structural support that is already visible and slow to reverse, such as the talent constraint and the divergence of capacity from maturity. Medium–High marks Calls with strong support but a residual timing or execution
risk. Medium marks Calls that are more likely than not but turn on a threshold (a "majority", a specific count) that could fall either way. Low–Medium marks Calls where the direction is right but the pace is genuinely uncertain, typically because the talent bottleneck constrains it.
The 2028 edition will open with the Scorecard. Each of the ten Calls will be graded hit, miss or partial, and a running hit-rate will be reported by confidence band, so that the calibration of the editors, not merely the direction of their judgement, is exposed to scrutiny. It is to be noted that a Call graded partial is a Call the method did not state sharply enough, and such outcomes are used to tighten the following year's drafting.
Every figure in this edition is drawn from a named third-party source and stamped with its year. Where sources disagree on scope, as they do on market size and capacity, a range is given rather than a false point estimate. Figures were current to July 2026.
The same year, read through four lenses (technology, markets, regulation and delivery) and then stress-tested against three scenarios and a watchlist.
The technology direction is consolidation and governance; the market direction is spend outpacing value until discipline arrives.
The technological through-line of 2027 is the movement from procurement to operation. The region has acquired the components of a modern data stack (hyperscaler regions, sovereign clouds, catalogue and governance tooling), and the work of the year is to integrate and operate them as platforms rather than to buy more of them. The arrival of the Microsoft Saudi Arabia East region at customer availability targeted for November 2026, on three availability zones with construction complete since December 2024, gives Saudi entities an in-region, sovereign-capable anchor for migration [Microsoft, 2024–26]. Expect consolidation of tooling, enforced classification carried through pipelines, and retrieval infrastructure built to feed models grounded context.
The market read is that spend continues to compound faster than value is realised. The GCC cloud market above thirty billion US dollars in 2025 and the data-centre market at 2.39 billion, compounding at 26.78 per cent to 2034, describe demand that is not in question [Mordor; IMARC, 2025]. What is in question is the return: projected cloud waste of 44.5 billion US dollars globally in 2025, and 29 per cent of infrastructure and platform spend wasted in 2026 [Flexera, State of the Cloud 2026] and a per-organisation data-quality cost near 12.9 million suggest that a growing share of the spend is not yet productive [Harness, 2025; Gartner via IBM, 2026]. The market correction of 2027 is not a fall in spend but the beginning of its measurement.
Regulation is tightening around classification and residency; delivery is shifting from the integrator to the in-house platform team.
The regulatory direction is towards demonstrable, classification-driven governance. Saudi Arabia's NDMO framework of fifteen domains, seventy-seven controls and 191 specifications sets a concrete compliance surface; Dubai's updated Data Manual of 1 July 2026 adds mandatory standards, self-assessment checklists and an AI-readiness scope; and Abu Dhabi's strategy commits to full sovereign-cloud adoption for government [NDMO, 2022; DDSE, 2026; DGE, 2025]. The notable gap is the UAE's Personal Data Protection Law, in force since 2 January 2022 but with executive regulations still unpublished in mid-2026, a source of engineering uncertainty for any federal-scope platform.
The delivery model is the domain in most visible flux. The integrator-led project, which has supplied most Gulf data capability to date, is giving way, slowly and at the leading entities first, to owned platforms operated by in-house teams. The Abu Dhabi Unified Government Data Centre for more than forty entities, announced with e& in October 2025, is the pattern in
physical form: build once, operate centrally, serve many [DGE, 2025]. The constraint on the shift is the talent bottleneck, which is why the corresponding Call is held only at Low–Medium confidence. Delivery, in short, knows where it must go and is limited by whom it can hire.
Technology consolidates, markets begin to meter, regulation demands demonstrable classification, and delivery moves in-house: all four pointing at the same foundation work the region has deferred.
The central case is that the foundation gap binds but begins to close at the leading entities; the tails turn on talent and on discipline.
Scenarios are Volge constructions built on the sourced figures in Parts I and III. They are illustrative reasoning about the year, not forecasts with assigned probabilities. Current to July 2026.
The signals to watch through the year: the seed of next year's Calls.
The Microsoft Saudi Arabia East region's general-availability date, and the migration volume that actually follows it through 2027. Publication of the UAE Personal Data Protection Law executive regulations, unresolved since the law took force on 2 January 2022. Production cross-entity data sharing at the SDAIA National Data Bank tier, Qatar's National Data Program and a UAE federal or emirate platform, the test of Call 1. Adoption of the updated Dubai Data Manual's AI-readiness scope and self-assessment checklists, launched 1 July 2026, across Dubai entities. Delivery of the Abu Dhabi Unified Government Data Centre for 40-plus entities, and whether the "AI-native government" 2027 target is met. Any public disclosure by a national entity of an AI failure or rescope attributed explicitly to data quality or engineering, the test of Call 6. The first named FinOps-for-data function or line item at a major Gulf entity, and evidence of tool-stack consolidation. Adoption of a data-mesh or data-product operating model at a Gulf anchor such as Aramco, ADNOC or a major bank. Middle East live IT capacity against the 1,669-megawatt 2026 forecast, and whether maturity metrics begin to close on the megawatts.
The 2028 edition opens with the Scorecard: each of these ten Calls graded hit, miss or partial, with a running hit-rate by confidence band. A forward-look that refuses to mark its own homework is just a brochure; Volge Quarterly marks it.
Reconciled from the sources listed. Ranges reflect genuine disagreement on scope and definition. Current to July 2026.
The operator's response to the year's foundation gap, and the reference matter (data appendix, sources and glossary) behind the narrative.
The numbers behind the narrative, in one place, each sourced and year-stamped.
| Metric | Value | Source & year |
|---|---|---|
| GCC cloud market, 2025 | >$30B (KSA $15.39B; UAE $12.84B) | Mordor Intelligence, 2025 |
| KSA cloud, 2025→2030 | $15.39B → $27.93B (12.66% CAGR) | Mordor Intelligence, 2025 |
| UAE cloud, 2025→2031 | $12.84B → $56.26B (27.93% CAGR) | Mordor Intelligence, 2025 |
| GCC data-centre market | $2.39B (2025) → $21.67B (2034), 26.78% | IMARC, 2025 |
| MEA colocation, 2025→2031 | $3.33B → $12.88B (25.32% CAGR) | Arizton, 2025 |
| ME live IT capacity | 761 → 1,115 → 1,669 MW (2024→26f) | Knight Frank / DC Byte, 2024–26 |
| UAE share of GCC capacity, 2025 | 35.2% | IMARC / Business Wire, Jan 2026 |
| GenAI pilots returning nothing | ~95% | MIT NANDA / Forbes, Oct 2025 |
| GenAI projects that fail | ~50% | Gartner, Jan 2026 |
| Agentic-AI projects cancelled by end-2027 | >40% | Gartner, Jun 2025 |
| Poor data quality cost, per org | ~$12.9M / year | Gartner via IBM, Jan 2026 |
| Projected cloud waste, 2025 | $44.5B | Harness, 2025 |
| Machine identities per human | 109 : 1 | Palo Alto Networks, 2026 |
| Gulf industrial power | $0.05–0.07 / kWh | Oxford Economics, Jan 2026 |
| PIF assets | ~SAR 3.4T at end-2025 [PIF annual report] | Sovereign disclosures, 2025–26 |
| AI to the Middle East by 2030 | ~$320B (KSA ~$135B; UAE ~$96B) | PwC, 2023–24 |
| SDAIA framework | 15 domains, 77 controls, 191 specifications | NDMO Standards, 2022 |
| Saudi National Data Bank | 360+ government systems integrated | SDAIA, 2025 |
Every figure in this edition traces to a named source below, stamped with its year. Figures were current to July 2026.
Definitions are Volge's, consistent with the national-programme and analyst sources cited throughout. Current to July 2026.
The terms this edition uses, defined for the operator and the board.
SDAIA: Saudi Data and Artificial Intelligence Authority; the national body housing the NIC, NDMO and NCAI. NDMO: National Data Management Office (Saudi Arabia); author of the 15-domain, 77-control, 191-specification data framework. National Data Bank: The Saudi national data lake integrating 360+ government systems. FCSC: Federal Competitiveness and Statistics Centre (UAE); operator of the federal open-data portal bayanat.ae. PDPL: Personal Data Protection Law; UAE Federal Decree-Law 45/2021, in force since 2 January 2022, executive regulations pending mid-2026. DDSE: Dubai Data and Statistics Establishment; the emirate data body under the Dubai Digital Authority. Dubai Pulse: Dubai's central data platform, delivering platform, data and infrastructure as a service. DGE: Department of Government Enablement (Abu Dhabi); absorbed the Abu Dhabi Digital Authority in 2023. NPC: National Planning Council (Qatar); established 2024, running the National Data Program. Data engineering: The discipline of ingesting, cleaning, cataloguing, governing and integrating data into a foundation usable by analytics and AI. Sovereignty by design: Architecting a platform so that classification-driven residency and access are enforced from the first line, not applied afterwards. FinOps: Financial operations applied to cloud and data spend; measuring, attributing and optimising the estate as a managed cost. Data mesh / data product: An operating model that treats data as an owned, served product per domain rather than a central monolith. Data classification: The scheme (e.g. Saudi Arabia's four levels) by which data is labelled to govern residency, access and handling. Agentic AI: AI systems that plan and act across multiple steps; noted by Gartner as facing a >40% cancellation rate by end-2027.
Volge Quarterly is Volge's signature series, published every quarter for enterprise customers. Each edition takes one subject, sets out where it currently stands on the published record, and makes a set of dated, falsifiable judgements on where it is going, written to give a buyer something firmer than a forecast: every figure carries its source and its date, every judgement carries the date on which it resolves, and the following year's edition grades each one hit, missed or partial.
Where a figure is not published, the edition says so rather than estimating it. Where published sources disagree, both are given and the disagreement is stated rather than averaged away. A number that does not survive checking is corrected here rather than repeated.
The edition carries no recommendation of any supplier and no offer of service. Where a vendor, platform or deployment is named it is because it is part of the published record of the subject.
This edition may be shared freely. It may not be altered or republished without permission, and figures carry their stated basis and date. Names and marks appearing in it are the property of their respective owners.