Everything About AI
NextFin News - Saudi Arabia and the United Arab Emirates are pushing ahead with data center plans worth tens of billions of dollars, but the gap between the gigawatts announced on stage and the megawatts actually drawing power is widening into the story's real risk. The kingdom's operational data center capacity stood at 467 megawatts in the first quarter of 2026, according to the Ministry of Communications and Information Technology, against a national target of about 1.5 gigawatts by 2030 - and the 6-gigawatt and 6.6-gigawatt figures in circulation belong to a single company's pipeline, not the country's grid.
The disconnect is the point. The Gulf is not short of ambition or capital. It is short of energized, tenant-confirmed capacity - and that gap is where the investment risk sits.
The Announcement Economy: Dollars Committed, Megawatts Pending
The pace of commitments is unmistakable. In Saudi Arabia, the second day of the LEAP 2026 technology conference on September 1 brought roughly $2.5 billion in new investments and launches across data centers, cloud services and technology localization, the Saudi Press Agency reported. Al Moammar Information Systems led with a $1.2 billion expansion aimed at lifting its total capacity to 192 megawatts; NHC Innovation announced an $880 million data center program in the Khuzam Digital Valley targeting 65 megawatts by 2033; Vision Invest and the Africa Finance Corporation committed $300 million to WIOCC for African digital infrastructure; and Mobily and BytePlus put more than $150 million behind cloud services.
Those announcements sit on top of earlier, larger commitments that define the scale of the ambition. HUMAIN, the artificial-intelligence company owned by the Public Investment Fund, and Blackstone-backed AirTrunk agreed in October 2025 to an initial investment of about $3 billion for a data center campus in the kingdom. DataVolt signed an agreement with NEOM for a $5 billion, 1.5-gigawatt AI factory in Oxagon, with a first phase expected to be operational by 2028. And in May 2026, HUMAIN tendered enabling infrastructure for a planned 6-gigawatt hyperscale AI campus in Riyadh - a package covering a 380kV/132kV/33kV electrical network, two substations rated at 500MVA and 200MVA, and a 2,000-megavolt-ampere bulk supply point.
Across the border, the UAE's anchor project is the 5-gigawatt UAE-US AI Campus in Abu Dhabi, unveiled in May 2025 and built by G42 with US partners under the US-UAE AI Acceleration Partnership. Spanning 10 square miles, it is designed to be the largest AI infrastructure project outside the United States, with compute resources reserved for US hyperscalers and approved cloud providers and served to nearly half the world's population within 3,200 kilometers.
Taken together, the GCC is tracking more than 174 major active and planned data center projects with a combined value above $93 billion. Technology spending across the Middle East and North Africa is forecast to reach $169 billion in 2026. The question is no longer whether capital is available. It is whether the power, security and delivery timelines can keep pace with the announcements.
Security Is Quietly Redrawing the Map
The geographic center of gravity is shifting, and the reason is not economics alone. In March 2026, drone strikes damaged two Amazon Web Services data centers in the UAE and a facility in Bahrain, causing structural damage, disrupted power delivery and fire-suppression water damage that took the sites offline. AWS said recovery would be prolonged given the nature of the physical damage, and customers including Careem, Alaan, Hubpay and Snowflake reported service disruptions.
That event did more than interrupt services for a few days. It forced a reassessment of a buildout model that had concentrated hyperscale capacity in coastal hubs optimized for submarine-cable latency. The emerging architecture favors strategic depth and sovereign capital able to underwrite hardened, distributed infrastructure - a profile that tilts toward Saudi Arabia's interior relative to the UAE's concentrated coastal footprint.
The shift shows up in the project pipeline. Saudi Arabia's largest announced campuses - HUMAIN's 6-gigawatt Riyadh site and DataVolt's Oxagon factory - are inland or purpose-built industrial zones with dedicated power arrangements. The UAE's 5-gigawatt campus, by contrast, is a government-to-government construct backed by explicit US security guarantees and know-your-customer protocols designed to prevent diversion of advanced semiconductors.
"Today's agreement is a testament to the ongoing collaboration between our countries in artificial intelligence," said Sheikh Tahnoon bin Zayed Al Nahyan, chairman of the Artificial Intelligence and Advanced Technology Council. "It is an expression of the UAE's commitment to pioneering innovation and fostering global collaboration in artificial intelligence, strengthening the UAE's position as a hub for cutting-edge research and sustainable development."
The security premium is now a line item in every Gulf data center pro forma. It is not merely an insurance cost; it is site selection, redundancy architecture, and the willingness of sovereign balance sheets to absorb risk that private developers cannot price.
Power, Not Chips, Is the Binding Constraint
Every announced gigawatt runs into the same wall: energization. A data center's capacity is the electrical power its equipment can draw, not its floor area or server count, and the gate that decides whether megawatts exist is the dated moment a utility connects the site to firm power. No Saudi AI project has published one.
The arithmetic is unforgiving. Saudi Arabia's operational capacity grew from 68 megawatts in 2021 to 440 megawatts in 2025 and 467 megawatts in the first quarter of 2026 - a sixfold increase over four years, on cumulative investment above SAR 16 billion ($4.3 billion) spread across more than 60 facilities, but still less than a third of the 1.5-gigawatt national target for 2030. HUMAIN's own interim target of 1.9 gigawatts by 2030, stated by chief executive Tareq Amin in November 2025, is 27 percent larger than the entire national 2030 target. Both cannot be delivered inside the decade without a step-change in grid-connection speed.
There is also a delivery-status gap inside the pipeline. HUMAIN broke ground on twin 100-megawatt campuses in Riyadh and Dammam in August 2025 with a committed go-live in the second quarter of 2026. As of the end of July 2026, no announcement, filing or credible report confirms either site is in commercial service, and the company has declined to disclose delivery progress when asked. The stc/center3 joint venture with HUMAIN, targeting up to 1 gigawatt, was extended but not signed as of a June 2026 filing. DataVolt's Riyadh East project - 44 megawatts in total, 36 megawatts anchored by HUMAIN - had not closed financing.
This is the mechanism behind the headline numbers: announcements move at the speed of press releases; energized capacity moves at the speed of substations, switchgear, water networks and grid interconnection. HUMAIN's 6-gigawatt campus entered the public record in May 2026 not with a compute order but with a procurement for electrical distribution and a bulk supply point - the clearest available statement of the binding constraint.
"Our strategic partnership with HUMAIN, a key player in the region, will support Saudi Arabia to realize its vision of being a data- and AI-driven economy," said Robin Khuda, founder and chief executive of AirTrunk. "We're bringing together the whole digital ecosystem, combining HUMAIN's end-to-end AI capabilities, from infrastructure to models, with AirTrunk's leading hyperscale data center capabilities."
The ecosystem is being built in the right order - infrastructure before compute - but that ordering guarantees a lag between commitment and capacity. Investors pricing Gulf AI exposure are effectively underwriting construction and grid-connection risk, not just demand risk.
The Cyclical-vs-Structural Call: A Regime Shift With a Delayed Delivery
This is a structural shift in where global AI compute is built, not a cyclical surge in data center spending that will mean-revert when the AI capex cycle cools. Three pieces of evidence separate the regime change from the cycle riding on top of it.
First, the driver is a permanent rewrite of the risk map, not a temporary supply-demand imbalance. Before March 2026, the dominant siting logic was latency to submarine cables and proximity to coastal enterprise hubs. After the strikes, physical survivability entered the site-selection model as a first-order variable. A risk model that has been rewritten does not revert on its own; it prices security permanently, the way term premia price duration risk after a debt-ceiling crisis.
Second, the capital structure is sovereign, not cyclical. A cyclical buildout is funded by private developers chasing a spread between contracted rents and construction costs - and it stops when that spread compresses. The Gulf buildout is funded by balance sheets executing national industrial strategy: Saudi Vision 2030, the UAE's AI and digital-economy programs, data-localization rules. Those mandates do not reverse when AI training demand pauses for a quarter.
Third, the chip-access architecture is institutionalized. The US-UAE framework hardwires American-managed cloud services into the region's most advanced campus, with know-your-customer protocols and diversion safeguards baked into the operating model. That is a structural moat, not a cyclical advantage.
The cyclical leg is real but distinct: the timing of announcements, the sequencing of groundbreakings, and the pace of tenant sign-ups will ebb and flow with AI demand, interest rates and equipment costs. If AI capex slows, some 2027-2028 go-live dates will slip. But the direction of capital - toward sovereign-backed, security-hardened Gulf campuses - is a regime shift. Getting this wrong flips the conclusion: a cyclical read says sell the announcements on any slowdown; a structural read says the announcements will arrive in waves over a decade, with the state absorbing the troughs.
The Second-Order Effect: Who Actually Bears the Risk
The first-order story is straightforward - more data centers are being built in the Gulf. The second-order effect is who ends up holding the risk when the megawatt gap persists. It is not the sovereign anchor. It is the private partner and, ultimately, the end customer.
Consider the transmission chain. An announcement commits capital at the headline level. The construction phase then exposes the developer to switchgear lead times, specialized labor, water infrastructure and grid interconnection - all of which have lengthened globally as AI campuses compete for the same constrained inputs. A one-year delay on a $3 billion campus does not merely defer revenue; it compounds financing costs and, in a sovereign partnership, typically renegotiates the risk-sharing toward the private side. The sovereign retains the strategic asset; the partner absorbs the execution variance.
The end customer faces a different second-order effect: concentration risk. As hyperscale capacity consolidates into a smaller number of sovereign-backed campuses, a single physical event - the March 2026 pattern - can disable a larger share of regional compute than the pre-2026 distributed model allowed. The hedge against that risk is precisely the distributed architecture the strikes made fashionable, but distributed edge nodes are cheaper to build and harder to harden. The region is likely to end up with a two-tier structure: hardened sovereign cores for critical workloads, and thinner edge capacity for latency-sensitive services. That is a more expensive architecture overall, and the cost will be priced into cloud contracts.
The third-order expectation gap sits in the equity market's framing of Gulf AI exposure. Investors underwriting data center developers, utilities and construction contractors are pricing a linear ramp from announcement to revenue. The infrastructure record suggests a step-function ramp instead - long flat periods of permitting and interconnection, punctuated by sudden energization events. Earnings models built on straight-line delivery will miss on timing even if they are right on total capacity.
The Counter-Thesis: Why the Gulf Still Wins the Race
The skeptical read - that announced gigawatts will not materialize - underestimates three structural advantages. First, capital is not the constraint. The Public Investment Fund held about $1.21 trillion in assets in 2025, and the kingdom announced a $40 billion AI investment fund in 2024; the UAE can draw on sovereign balance sheets with similar depth. When the binding constraint is capital-intensive grid infrastructure, sovereign capital is the one input the Gulf has in abundance.
Second, the power constraint cuts both ways. The UAE's 5-gigawatt campus is designed to draw on nuclear, solar and gas generation - an energy mix the Emirates can assemble faster than most jurisdictions. Saudi Arabia's advantage is cheaper marginal power and land, which matters for AI training workloads that are power-cost sensitive rather than latency sensitive.
Third, the security realignment is not a retreat from the UAE; it is a diversification. The US-UAE framework explicitly reserves advanced compute for American-managed cloud services, giving the Emirates a privileged channel to chips that remain restricted elsewhere in the region. That is a moat no inland site can replicate with capital alone.
The strongest version of the counter-thesis, then, is not that the Gulf buildout will fail. It is that the region will deliver - but with a different geography, a longer timeline, and a higher share of sovereign ownership than the announcements imply. Private developers and international capital will participate - Blackstone, through AirTrunk, is proof - but the risk-bearing layer will sit with the state.
"We are thrilled to help power this next era of innovation in the Middle East and enable AirTrunk's expansion in this important region," said Stephen A. Schwarzman, chairman and chief executive of Blackstone. "The AI revolution continues to be one of Blackstone's highest conviction themes, and we bring scale and expertise across the ecosystem as the largest provider of data centers globally and a significant investor in related services and infrastructure."
What to Watch: The Signals That Separate Hype From Capacity
The forward look splits by horizon. In the short term - the next two to three quarters - the signal is energization, not announcements. A dated utility interconnection for HUMAIN's Riyadh or Dammam campuses, or a financing close on DataVolt's Riyadh East package, would convert pipeline into capacity. Absent those, the 467-megawatt baseline will drift upward only gradually.
Over the medium term - 2027 to 2028 - the test is tenant confirmation. A hyperscaler or frontier-model developer disclosing production workloads on Saudi or UAE capacity would settle more than any developer statement. The NEOM DataVolt phase, targeted for 2028, is the first large-scale proof point on the calendar.
Over the long term, the structural question is whether the Gulf becomes a net exporter of compute or a sovereign-cost center. The base case is a hybrid: latency-insensitive training workloads migrate to the Gulf's cheap power, while latency-sensitive inference and regulated workloads stay closer to end markets. The upside case is that US security frameworks turn the UAE into the sanctioned-world's gateway to advanced compute for nearly half the global population within 3,200 kilometers. The downside case is that grid-connection timelines stretch announced 2026-2027 go-lives into 2029-2030, stranding capital in partially built campuses.
The falsifying signal for the cautious view is specific: if Saudi Arabia's operational capacity prints above 900 megawatts by the end of 2027 - roughly double the first-quarter 2026 level - the delivery-gap thesis is wrong and the buildout is executing faster than the infrastructure record suggests. Conversely, if capacity remains below 600 megawatts at end-2027 while announcements continue to accumulate, the megawatt gap is structural, not cyclical.
Bottom Line
Saudi Arabia and the UAE are not short of ambition or capital - the $93 billion GCC pipeline proves that. What they are short of is energized, tenant-confirmed capacity, and that gap is where the risk sits. The March 2026 strikes accelerated a security-driven realignment that favors Saudi Arabia's strategic depth, but power infrastructure, not press releases, will decide who actually hosts the AI workloads of the next decade. Watch the substations, not the stage.
Explore more exclusive insights at nextfin.ai.
