Across the GCC, public reporting tracks more than 174 projects worth over USD 93 billion. The UAE runs the region’s most connected AI platform; Saudi Arabia is assembling its largest state-backed compute economy. Both ambitions route through one strait, one set of imported chips, and one neighbourhood at war.
The Gulf is two strategies told at once. The UAE is building the region’s most globally connected AI and cloud platform, plugged into US chips and hyperscaler capital. Saudi Arabia is building the region’s largest state-backed domestic compute economy, anchored to Vision 2030 and sovereign capital. Both treat data centres as instruments of power, not real estate.
On the UAE side, the marquee line is Stargate UAE, a 1 GW named cluster inside a 5 GW UAE–US AI Campus envelope, with USD 8 to 10 billion of investment and roughly 200 MW targeted by 2026 to 2027, alongside Microsoft’s USD 7.9 billion UAE commitment for 2026 to 2029, part of a USD 15.2 billion total across 2023 to 2029, delivered largely through Khazna, and a 35,000 advanced-chip authorisation for G42. On the Saudi side, HUMAIN sits at the centre of a state buildout: an AirTrunk partnership, AWS’s USD 5.3 billion region (part of an AWS commitment now above USD 10 billion in the kingdom), a USD 10 billion Google and PIF hub, and a USD 6 billion PIF data-centre programme. Since then the stack has deepened: a USD 10 billion, 500 MW compute deal with AMD, first racks from Qualcomm and Cisco, and joint ventures with STC (1 GW) and xAI (500 MW), all pointed at a ten-year target of 6.6 GW across 211 secured land plots.
Set against demand, those commitments look outsized. The data-centre market the build is meant to serve is forecast to roughly triple by 2030: Saudi Arabia from about USD 1.33 billion in 2024 to USD 3.9 billion, and the wider GCC from roughly USD 3.5 billion to USD 9.5 billion. That market revenue is a fraction of the capital pipeline above, which is the tell: the Gulf is provisioning compute ahead of its own demand, wagering on regional and export workloads it intends to capture rather than ones it already has.
What makes this chapter different from the others in the series is the risk layer. Once semiconductors enter the story, the model depends on secure shipping lanes, imported hardware and strategic materials. The Iran war has already moved that risk from abstract to physical: the first wartime drone strikes in history deliberately aimed at commercial data centres, hitting AWS sites in the UAE and Bahrain, helium and bromine exposure through the Strait of Hormuz, and renewed scrutiny of every hyperscaler commitment in the region.
Four instruments read the Gulf: The Exposure, a scenario-modelled file on what regional war does to investment and viability; a threat map of compute, bases and targets; The Brief on how Saudi megaproject ambition has been redrawn over time; and a comparison of the two national models. Every modelled figure is tagged, and every projection is framed as such.
Once semiconductors enter the model, the Gulf’s build depends on secure shipping lanes, imported hardware and strategic materials, all routed through the world’s most militarised water. Five files map where that exposure sits. The production desk then models it: what war does to the build, who captures the capital that flees, and how the region is hedging and de-escalating.
An interactive model in four layers: what conflict does to investment and viability, who captures the capital that flees the Gulf, how the region hedges and de-escalates, and the incident record behind it.
A directional Entelligencia model, not a forecast. Pick a scenario and intensity, then read four layers: what it does to the build, who captures the diverted capital, how the region mitigates, and the incident record. Figures are ranges and directions, not precise outcomes.
The Gulf’s data centres sit beside the world’s densest concentration of US military infrastructure, the region’s oil and gas spine, and a strait that roughly a quarter of the world’s seaborne oil moves through. Each node is a real site, colour-coded as a compute cluster, a US base or an energy target. Open one for detail; toggle the overlays for the Hormuz chokepoint, cross-Gulf strike reach and the subsea cables. Positions are approximate: a strategic schematic, not a survey.
Sixteen real sites plotted against the US military footprint, the energy spine and the Strait of Hormuz. Open any node, filter by country, and toggle the chokepoint, strike-reach and subsea-cable overlays.
The load-bearing numbers behind the region’s compute push, each graded against the record and sourced. Chip access has been granted, the capital is committed, and the strait has closed, partly reopened and closed again inside six weeks. This is the part of the chapter that moves.
The live layer of the chapter: the load-bearing numbers behind the Gulf’s compute push, each graded against the record and sourced. This is the part that moves, and it is maintained as it does.
10 July 2026: US Commerce moved the UAE into Country Group A:5, widening eligibility for License Exception STA on specified items. BIS separately identified G42 and Core42 for specified advanced-computing authorisations. The two are different instruments and are easy to run together: A:5 is a country classification, not blanket licence-free access for every UAE state or investment vehicle, and the recipient-specific authorisation does not name MGX.
Functionally shut as at 1 August 2026 per Iranian state media’s declared suspension, a claim independent maritime trackers have not confirmed outright. The single status understates the month: a US and Iran memorandum signed 14 June extended the ceasefire and provided for a phased reopening, transits recovered to nearly 400 vessels in the week of 20 June, and the ceasefire then collapsed on 11 to 12 July. Closure, partial reopening and re-closure inside six weeks. We publish the status and the volatility together.
Abu Dhabi, with G42, OpenAI, Oracle, Nvidia, Cisco and SoftBank. The three figures in circulation are nested, not additive: a 5 GW UAE–US AI Campus envelope, a 1 GW Stargate UAE cluster inside it, and a 200 MW first phase inside that, due online in 2026. Powered by Barakah, solar and gas. Where a Gulf number appears without a basis, this is usually the error underneath it.
The Abu Dhabi state AI vehicle’s first fund, raised to back compute and model builders across the stack.
Announced 28 April 2026 and effective 1 May, after 59 years. The exit is verified; reading it as a pivot from oil quotas toward compute as the strategic export is this desk’s analysis, not a stated reason.
Four APR-1400 reactors supplying roughly a quarter of UAE electricity, the carbon-free baseload underwriting the campuses.
With HUMAIN, up to 150,000 GPUs targeted, part of the Kingdom’s hyperscaler build-out.
A $5bn green data campus at NEOM, with a first 300 MW phase targeted by 2028.
HUMAIN’s flagship government supercomputing campus, anchoring Saudi sovereign compute.
Committed to AI infrastructure worldwide since 2025, with Gulf vehicles, MGX, PIF, QIA, ADQ, behind roughly two-thirds.
PIF’s data-centre capacity goal for the Kingdom by 2030, part of a plan to draw 12% of GDP from AI.
The ten-year Emirati investment commitment into US technology and infrastructure underpinning the chip deal.
PortWatch recorded just 10 transits of the strait on 23 July 2026 against an 88-per-day pre-crisis baseline, and USNI News reports a 52.4 per cent collapse between 20 July and the end of the month. The severity is corroborated even where the declared closure is not.
As at 1 August 2026. An oil spill from a sanctioned shadow-fleet tanker off Oman has added an environmental dimension to a chokepoint already carrying roughly a quarter of the world’s seaborne oil.
March 2026: confirmed drone strikes hit two AWS facilities in the UAE and damaged one in Bahrain. The damage, the regional disruption, the multi-month repair outlook and the waiver of all March 2026 usage charges are all verified. The roughly USD 150m figure attached to them is a third-party estimate, not a cost AWS has disclosed, and the headline is graded to the weakest element. AWS strongly recommended customers migrate to other regions and hold remote backups; it did not tell them the move was permanent.
Grid capacity accessible behind HUMAIN’s land position, secured in six weeks after chief executive Tareq Amin asked sixteen government entities to identify suitable substations. This chapter records the plots and the 6.6 GW ten-year target; the power behind them is the part that was missing.
A campus tendered across 24 square kilometres. The Saudi build is no longer constrained by finding sites, which is the constraint most other markets in this report are still solving for.
HUMAIN mandated Goldman Sachs in May 2026 to advise on a data-centre financing package worth at least 20 billion riyals, about USD 5.3bn, aimed at 2 GW of capacity. A USD 1.2bn facility from the National Infrastructure Fund, announced at Davos in January 2026, preceded it. The build is moving from announcement to structured finance.
A July 2026 HUMAIN collaboration with Canada’s Cohere for dedicated AI compute, live by Q4 2027. Small in megawatts and large in signal: the sovereign-compute model is actively diversifying beyond US hyperscalers.
HUMAIN 51 per cent and STC 49 per cent, targeting 250 MW initially and scaling toward 1 GW. July 2026 reporting frames the arrangement as a memorandum with a six-month term rather than a binding, fully financed structure. The 1 GW this chapter records is the target of that structure, not a financed commitment, and the distance between the two is the point.
The February 2026 Series E investment, since converted into SpaceX shares after SpaceX absorbed xAI. A sovereign compute champion now holds equity in a launch company.
This chapter previously carried Riyadh and Dammam as live in Q2 2026. Construction at both is verified and Q2 2026 is a verified stated target, but as at 31 August 2026 no operator announcement, customer confirmation, utility energisation record or regulatory filing independently confirms both 100 MW sites in commercial service. Physical development: verified. Full 200 MW delivered: not verified. Separately and later, HUMAIN awarded MIS work covering a further 200 MW on top of an existing 50 MW project, taking that contract scope to 250 MW – awarded construction, not delivered capacity – and an AMD and Cisco announcement placed production AI compute live in the Kingdom, which validates a deployment without establishing two complete campuses.
A practising partner on what actually decides a Gulf data-centre deal. His answers do not all run with the grain of the questions we put to him: he reframes the question of regulatory mismatch, he resists the idea that contracts are routinely unfit for AI, and he rejects the notion of one dominant procurement structure. His positions are reported here as he gave them, summarised in our house style rather than quoted verbatim. Open any file for his response.
The Gulf inverts the constraint this report tracks everywhere else. In Europe the story is scarcity, waiting years for a connection and bending the programme around when power arrives. Here operators are building the energy themselves, at gigawatt scale, and that pulls the whole energy chain inside the scope of what has to be proven. What is being assured is no longer a facility that draws power, but generation, storage, protection and distribution across a campus expected to stand on its own.
Where Europe waits for a connection, Gulf operators build their own generation. The commissioning scope changes materially: verifying how a campus behaves when islanded or switching between sources, rather than confirming a utility feed is live.
Cooling has to reject heat at ambient temperatures that would be an extreme event elsewhere and are simply summer here, with dust and humidity affecting both the equipment and the testing itself.
Mobilising specialists into the region works in the near term and cannot be the long-term model. The adaptation is pairing that with a training pipeline so the standard becomes rooted rather than flown in.
Frameworks around privately owned high voltage and sovereign data are still maturing, so protocols are developed project by project from international practice, and authorities engaged early.
Contributed to The Next Hotspot via the Entelligencia briefing survey, attributed by name with firm and title at the contributor’s request. Views attributed to the author. His full read on lifecycle assurance is published as Mind the Gap.
A power and digital-infrastructure veteran of a decade across the Gulf and APAC, contributing in a personal capacity. His read runs underneath the capacity tables: why a market that matured in two years cannot be planned for, why geopolitics rather than capability picks the supplier, and which MEA market takes off next.
A project director’s read, contributed to The Next Hotspot
This chapter has counted megawatts, capital and political risk. It has not priced the handover. Muhammad Bilal has run Gulf builds through to live, including a hyperscale-ready facility in Oman, and his argument is that the region’s schedules do not slip on concrete. They slip in the weeks when a finished building has to prove it works as one system, in 50°C heat, with something already broken. Seven files on the part of the build the announcements never mention.

Commissioning is his trade, which is worth knowing while reading a case for commissioning.
It is also why the argument is worth testing rather than taking on trust, and his claims are testable without relying on his firm at all. The ambient figures are matters of record. The distinction between component sign-off and integrated systems testing is an engineering fact, not an opinion. And his talent argument is the same finding The Talent Crunch reaches from a different direction entirely, which is the kind of agreement worth more than either source alone.
Where he is making a claim about frequency rather than mechanism, that a facility is handed over unready “more often than the industry admits”, it is graded Contributed and stays there. No one publishes that number.
Contributed via the Entelligencia briefing survey, 28 July 2026, and published with the contributor’s approval. Seven questions were put; all seven were answered. Quotations are his own words, unedited. Where his answers have been compressed into summary above, the summary is ours and is marked as such rather than set inside quotation marks. Firm and title confirmed by the contributor on 29 July 2026.
The Gulf sells the render before it builds the thing. Saudi Arabia in particular unveils world-redefining visuals, then quietly redraws them as cost and capacity bite. That gap between projection and build is the single most useful lens on any Gulf compute announcement. Scroll each track left to right: the hologram fades into what stands today. Scaled-back figures are contested and framed as reported, not settled.
The same instinct that produced a 170km mirrored city and a fifteen-stadium tournament now produces gigawatt AI campuses and sovereign-chip pledges. The ambition is genuine and the capital is real, but the delivery curve in the Gulf has a consistent shape: a maximal reveal, a quiet rescoping, then a smaller but still substantial build. For an investor or operator, the discipline is to separate the announced number from the financed one, exactly as the dashboard does for capacity. The render is the marketing; the megawatt is the asset.
Six markets, one buildout. This is the full comparative file behind the chapter: the data-centre base, the economy, power, people, resources, institutions, the operators on the ground and the resilience to absorb a shock. Choose who to compare, then read each factor by the figures and the long view side by side.
Each spoke is one of the eight dimensions, scored 0 to 10 for like-for-like comparison – indicative editorial weighting, not a published index. The bars rank the composite. Click any spoke, or a tile below, to open that factor. Toggle a market to add or remove its shape.
Five named figures across sovereign-AI champions, international capital and national policy. Each card opens to a public-record profile: the seat they hold, the institutions behind them, and where their mandate meets the compute build. Roles and figures are drawn from public company and government records.





A market is not settled on a spreadsheet. It is argued out by the financiers, bankers and platform-builders who decide where the next megawatt and the next rack land. Each seat opens to their file and a clip from the conversation. Audio here is representative; final recorded remarks, a contributor note or a podcast segment drop into the same player.
Sovereign capital sets the mandate; suppliers and the power layer gate what can actually be built; the operators raise the halls that the hyperscalers and the sovereign-AI programmes anchor. The Gulf signature is that the same funds often own the capital, the operator and the demand, so the chain is shorter and more state-directed than anywhere else in the report. The clearest exception is KKR’s more than USD 5 billion behind Gulf Data Hub: global private capital underwriting an independent operator rather than a national champion, the one tier where the chain runs on outside money instead of the state’s.
05 tiers · 30 entitiesLogos will be added later and will identify the parties for evaluation only, not as an endorsement, subject to each owner’s brand-usage rights before any commercial use. Power and supplier names are indicative of the categories active across the GCC. Connections are Entelligencia’s reading of public deal and partnership announcements, and is tagged as a reading.
Where the build actually happens, and what is in the way.


