An audit of Vision 2030’s 1.5 GW target against HUMAIN’s 6.6 GW pipeline, the SEC interconnection queue and the substations that decide delivery
By: Logan Pierce – SeaPRwire – Saudi Arabia operated 467 megawatts of data center load in the first quarter of 2026. That number, published by the Ministry of Communications and Information Technology, is the one that matters, and it is the number that almost never appears in the headlines. It is up from 440 megawatts at the end of 2025 and from 68 megawatts in 2021, a sixfold increase in four years built on cumulative investment above SAR 16 billion, roughly US$4.3 billion, spread across more than 60 facilities. It is also less than a third of the Kingdom’s own 2030 national target of 1.5 gigawatts, and about one-fourteenth of the 6.6 gigawatt pipeline that HUMAIN, the Public Investment Fund’s AI company, says it intends to deliver by 2034.
The gap between those two registers is the story. One is measured in energized load, the kind that a utility has physically connected to firm power and that a customer can bill against. The other is measured in press releases, memoranda of understanding and procurement notices for electrical distribution equipment. Both are real. Only one of them can train a model. This article is an attempt to keep the two apart, to show where the pipeline is genuinely deep, where it is thin, and which specific gates will determine how much of the announced capacity is live by the end of the decade.
Start with what is energized, not what is signed
The most useful discipline when reading Gulf AI infrastructure news is to ask a single question of every number: is this connected? Applied to Saudi Arabia, the question collapses a confusing pile of announcements into a short and legible list.
| Milestone | Figure | Status | Source |
| Operational load, 2021 | 68 MW | Actual | MCIT via SPA, 2026-08-15 |
| Operational load, end-2025 | 440 MW | Actual | MCIT via SPA, 2026-08-15 |
| Operational load, Q1 2026 | 467 MW | Actual | MCIT via SPA, 2026-08-15 |
| Facilities in operation | 60+ | Actual | MCIT via SPA, 2026-08-15 |
| Installed IT load, mid-2026 | ~410 MW across ~40 facilities | Estimate | Alvarez & Marsal, 2026-08-09 |
| Cumulative investment | SAR 16 billion (US$4.3 billion) | Actual | MCIT via SPA, 2026-08-15 |
| National target, 2030 | 1,500 MW | Target | National Data Center Strategy, June 2025 |
| HUMAIN medium-term guidance | 1,900 MW | Target | Tareq Amin, November 2025 |
| Announced AI-grade pipeline | 6,600 MW | Announced | Multiple, low verification |
| Planned capacity, all projects | 3,400 MW | Target | Asharq News via MOFCOM, 2026-06-04 |
Two rows in that table are worth dwelling on because they do not agree with each other, and the disagreement is not an error. The MCIT figure of 467 megawatts counts operational load. The Alvarez & Marsal figure of roughly 410 megawatts across about 40 facilities counts installed IT load in operating facilities, and it was published a week earlier. Different denominators, different cut-off dates, different definitions of what counts as a facility. Both are cited in this article rather than reconciled, because the reconciliation has not been published by anyone with access to the underlying data. A reader who sees 467 and 410 in the same week should conclude that the Kingdom’s live capacity is somewhere in the low hundreds of megawatts, not that one source is wrong.
The second useful observation is the shape of the growth curve. The jump from 68 to 440 megawatts happened over four years. Getting from 467 megawatts to 1.5 gigawatts by 2030 requires adding roughly three times as much capacity again, in half the time, at densities that are an order of magnitude higher per rack. That is not impossible. It is simply a different kind of project than the one the Kingdom has already executed, and the difference is concentrated almost entirely in the electrical infrastructure rather than in the buildings.
Four clusters, four sets of economics
Saudi Arabia is not one data center market. Treating it as one is the most common analytical error in coverage of the region, because the four clusters have different power costs, different water economics, different permitting environments and different customers.
- Riyadh metropolitan anchors enterprise, government and SDAIA-adjacent workloads. It has the deepest fibre and the shallowest pool of available large-block contiguous power. It is also where the Hexagon facility sits, the SDAIA program reported at roughly 480 megawatts and designed to Tier IV standard, which would make it the largest government data center in the world on completion.
- Eastern Province, around Dammam and Jubail, is the cheap-power, heavy-industry profile. It sits on top of the gas backbone and close to Aramco’s industrial load. HUMAIN broke ground on a 100 megawatt campus here in August 2025 alongside a twin campus in Riyadh.
- NEOM and the northwest is the moonshot. DataVolt’s program here is the single largest announced build in the Kingdom at 1.5 gigawatts of contracted IT load, tied to a multi-decade power purchase arrangement and scoped as renewables-first from the outset rather than retrofitted onto a gas posture. The 100 megawatt HUMAIN and DataVolt joint facility at Oxagon broke ground in August 2025 with a stated 2028 commercial operation date.
- The Red Sea and western corridor, Jeddah, Yanbu and King Abdullah Economic City, handles connectivity, content delivery and a growing share of hyperscaler edge, largely because that is where the submarine cable landings are.
The practical consequence is that a customer evaluating Saudi capacity should not ask “how much does a megawatt cost in Saudi Arabia.” They should ask which cluster, at which density, under which power purchase arrangement, and with which water source. The answers diverge by enough to change the economics of a deployment.
The binding constraint is not chips
Every announced gigawatt in the Kingdom runs into the same wall, and it is not export controls. It is energization. A data center’s capacity is the electrical power its equipment can draw, and the gate that decides whether those megawatts exist is the date on which a utility physically connects the site to firm power. No Saudi AI project has published that date.
The arithmetic of the power system makes the problem concrete. Saudi Electricity Company delivered 160.5 terawatt-hours to grid customers in the first half of 2025, an annualized rate near 321 terawatt-hours. Against that base, the national 1.5 gigawatt target for 2030 represents roughly 3.7% of current grid deliveries. HUMAIN’s 6.6 gigawatt pipeline represents roughly 16%. Those two figures describe completely different engineering problems. The first can plausibly be absorbed by a system that is already growing and already building generation. The second implies dedicated generation, dedicated transmission and a connection queue that does not currently exist, and it implies all of it inside eight years.
The equipment lead times compound the schedule risk in a way that is easy to underestimate. Transformers carry two-to-four-year lead times in the current market. Switchgear at the ratings that Blackwell-class deployments require is in the same position. SEC interconnection studies and power purchase agreements typically run six to nine months for sites under 100 megawatts, and longer for the 50-to-200 megawatt loadouts that AI campuses actually need. A twelve-to-eighteen-month slip in grid connection does not merely delay a project. On a market whose entire competitive advantage is cheap power, it erases the tariff advantage that justified siting there in the first place.
The clearest evidence for where the constraint sits is procedural rather than rhetorical. HUMAIN’s 6 gigawatt campus entered the public record in May 2026 not with a compute order but with a procurement for electrical distribution equipment and a bulk supply point. Tenders in the first half of 2026 covered a 2,000 MVA bulk supply point and two substations. When a company that already has capital, land and a sovereign sponsor goes to market first for substations, it is telling you what it believes the bottleneck is.
Where the pipeline is thin
Announcements move at the speed of a communications department. Energized capacity moves at the speed of substations, switchgear, water networks and grid interconnection. The gap between those two speeds shows up in the delivery record, and the delivery record in Saudi Arabia currently has several visible soft spots.
- 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 that either site has entered commercial service. The company has declined to disclose delivery progress when asked.
- The stc and 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 with 36 megawatts anchored by HUMAIN, had not closed financing.
- The 1.9 gigawatt medium-term guidance from HUMAIN’s chief executive is 27% larger than the entire national 2030 target. Both cannot be delivered inside the decade without a step change in grid-connection speed, and no one has explained how the two figures are meant to coexist.
None of this means the projects are cancelled. It means that the honest way to read a Saudi data center announcement is as a statement of intent with an unspecified delivery date, and that the date is set by a utility rather than by a developer. Investors pricing Gulf AI exposure are underwriting construction and grid-connection risk, not only demand risk.
There is also a definitional trap worth naming. Several trackers report “up to 1 gigawatt” or “1.9 gigawatts” as if these were operating capacity. They are not. They are corporate pipelines measured against a national target that is itself a target. A figure that has not been energized should never be presented as capacity that exists, and the discipline of separating the two is the single biggest improvement most coverage of this market could make.
Who is actually building
The vendor layer in the Kingdom is genuinely deep, which is the strongest argument that the buildout is structural rather than cyclical.
| Builder | Commitment | Status |
| AWS | US$5.3 billion, Saudi region | Operational, opened 2026 |
| DataVolt | US$5 billion, NEOM Oxagon, 1.5 GW | Phase 1 100 MW under construction, 2028 target |
| AirTrunk and HUMAIN | ~US$3 billion initial investment | Partnership announced |
| HUMAIN, AMD, Cisco | Joint venture, 250 MW phase from 2027 | Announced |
| SDAIA Hexagon | 480 MW, Tier IV, Riyadh | Groundbreaking January 2026 |
| Oracle Cloud Infrastructure | Riyadh region | Operational |
| Google Cloud | Saudi region plus HUMAIN collaboration | Committed |
| Microsoft Azure | Organic build plus G42 stake | Committed |
| Alibaba Cloud | Regional node | Operational |
| Center3 (stc) | 1 GW colocation by 2030 | Committed |
Two things follow from that table. First, the hyperscalers are present and operating, which means the control-plane, compliance and availability-zone work is being done and will outlast any single AI capex cycle. Second, the largest single program inside the Kingdom is HUMAIN’s, and it is the one with the least published delivery evidence. That asymmetry should be visible in any risk assessment of the market.
The licensing layer underneath all of this is more developed than outsiders expect. The Communications, Space and Technology Commission, formerly CITC, established a data center service provider registration regime under its Regulatory Framework for Data Center Service Provision, covering both wholesale and retail operators providing data center services to third parties inside the Kingdom. Registration is classified by facility development stage, with a Qualifying category reserved specifically for data centers under construction, and each facility must be registered individually by classification and development stage. The public register lists 69 registered data center service providers. That is a real regulatory apparatus, and it is a reason to take the market’s institutional depth seriously even where individual projects slip.
The power mix question nobody has answered
Saudi Arabia’s competitive advantage in AI compute rests on three things that are genuinely hard to replicate: power in the US$20 to US$50 per megawatt-hour range, sovereign capital willing to absorb construction risk that private developers cannot price, and connectivity across three continents. Chips can be export-controlled, models can be leapfrogged, and talent can be poached. Cheap firm power next to submarine cable landings cannot be reproduced by policy in Virginia, Ireland or Singapore, where grid stress already constrains AI growth.
What the advantage does not guarantee is the renewable share. The honest assessment of the Saudi pipeline is that the renewable buildout is running behind the pace needed to power AI compute at full scale by 2030. Grid infrastructure, storage deployment and permitting remain bottlenecks even as flagship solar projects proceed, which means natural gas will carry a large majority of the cumulative megawatt-hours consumed by Saudi AI data centers through the 2025 to 2028 period. Realistic estimates put the renewable share of Saudi data center consumption at 30% to 50% by the end of the decade.
That matters commercially, not only environmentally. Hyperscalers buying Saudi capacity are buying against internal scope 2 commitments, and a campus that is 70% gas-fired in 2028 will be priced and contracted differently from one that is 70% renewable. The Kingdom has two partial answers. NEOM and DataVolt were scoped as renewables-first, with a power purchase arrangement tied to NEOM’s renewable and green hydrogen architecture, so their energy mix and power usage effectiveness profile will diverge from the rest of the Saudi fleet over time. And the regional precedent is moving: the Masdar and EWEC round-the-clock renewable project in the UAE, which closed US$6.1 billion of financing in July 2026 for 5.2 gigawatts of solar paired with 19 gigawatt-hours of battery storage to deliver 1 gigawatt of baseload, is the first demonstration that firm renewable power at this scale is commercially financeable rather than a pilot.
How the Kingdom compares with its nearest rival
The comparison that matters is not with Virginia or Frankfurt. It is with the United Arab Emirates, which is running the same playbook on a shorter clock and with a head start of roughly a decade. The UAE’s advantage is that its operating base is larger and its interconnection process is more standardized; EWEC’s own analysis expects UAE data center electricity consumption to exceed 6 terawatt-hours by 2030. Saudi Arabia’s advantage is land, cheap gas and a sovereign balance sheet that can pre-fund generation and transmission rather than waiting for a tariff to justify it.
The UAE’s answer to the firm-power problem is further along in one specific respect. The Masdar and EWEC round-the-clock project pairs generation with 19 gigawatt-hours of storage to deliver a flat 1 gigawatt, and it closed financing in July 2026. Saudi Arabia has nothing public at that scale, and the gap between the two countries on firm renewable capacity is currently wider than the gap between them on announced IT load. That inversion is worth noting, because it means the Kingdom’s headline lead in gigawatts rests on an energy buildout that is behind its neighbour’s.
Where Saudi Arabia is clearly ahead is institutional scaffolding. The CST registration regime, the National Data Center Strategy published in June 2025 under MCIT and SDAIA, and the explicit Tier III and Tier IV classification of government facilities give the market a rulebook that the UAE has not published in comparable detail. For a tenant choosing between the two, the trade is between a market with more operating capacity today and a market with clearer rules for building tomorrow.
What to ask before you sign anything
The questions below are the ones that separate a deliverable contract from a hopeful one, and they are ordered by the sequence in which the answers become unavailable.
- What is the energization date, and who carries the risk if it slips? A construction milestone is not an energization milestone. If the contract does not distinguish them, the tenant is underwriting the utility’s schedule.
- Is the power purchase arrangement signed, or is it a letter of intent? Several Saudi projects have procurement activity without closed financing. A PPA that is not signed is not a PPA.
- Which cluster, and what is the water source? Liquid-to-the-chip cooling at Blackwell-class densities changes the water economics materially, and the answer differs between Riyadh, Dammam, NEOM and Jeddah.
- What is the renewable share for this campus in year one, and what is the contracted path? If the answer is a national average rather than a campus figure, treat it as zero.
- What happens to the tariff if interconnection slips twelve months? This is the clause that decides whether the cheap-power advantage survives a delay, and it is the clause most often left unwritten.
Three indicators that will settle this
The Kingdom has proven it can turn capital into operational data centers. Whether it can turn them into a durable export industry is a question the 2026 to 2030 window will answer, and three published indicators will answer it early.
- The renewable share of data center consumption, disclosed per campus rather than nationally. If that number climbs above 50% before 2030, the energy advantage is compounding. If it stalls near 30%, the gas dependency becomes a contracting problem for every customer with a scope 2 target.
- Published PPA pricing for AI-grade load. Today almost nothing is public. The first credible published tariff for a firm, high-density connection in the Kingdom will tell you more about the market’s real cost base than any capacity announcement.
- The SEC transmission cadence, measured as the interval between energization dates rather than between groundbreaking dates. Two or three campuses going live within a quarter of each other in 2027 is the signal that the constraint has been broken. A steady stream of groundbreakings with no energizations is the signal that it has not.
There is a fourth, softer one: whether HUMAIN publishes an energization date for the Riyadh and Dammam campuses it committed to the second quarter of 2026. It has not, and the absence is more informative than any statement it has issued.
What the 1.5 gigawatts is actually worth
Set the whole pipeline aside for a moment and consider just the national target, because it is the number with a government behind it. 1.5 gigawatts by 2030 is about 3.7% of current grid deliveries. It is roughly three times the capacity that exists today. It is achievable on the engineering, and the constraint is scheduling rather than physics.
Against that, the announced pipeline of 6.6 gigawatts is roughly 16% of current grid deliveries and roughly fourteen times operating capacity. It is not achievable on current interconnection cadence, and no serious analyst should model it as if it were. The interesting question is not whether Saudi Arabia will materially scale its compute footprint, because it will. It is how much of the announced pipeline energizes on the disclosed timetables, and the answer to that will be written by substation delivery schedules rather than by anyone in Riyadh or Washington.
For anyone allocating capital or engineering time against this market, the operational conclusion is mundane and unglamorous. Site selection should be driven by published grid connection dates, not by announced capacity. Contract structures should tie payments to energization milestones rather than to construction milestones. And any model that treats announced gigawatts as addressable capacity before 2030 is modelling a press release, not a power system.
Sources
- Saudi Press Agency via MCIT, operational data center capacity 68 MW (2021), 440 MW (2025) and 467 MW (Q1 2026), 2026-08-15 — https://www.spa.gov.sa/fa/N2654085
- Alvarez and Marsal, Financing Saudi Arabia’s Data Center Buildout, 2026-08-09 (via MOFCOM) — https://sa.mofcom.gov.cn/ztdy/art/2026/art_657744ccb82140649c8a34a132afd4ac.html
- Communications, Space and Technology Commission, Data Center Service Provision regulatory framework and public register of 69 providers — https://www.cst.gov.sa/en/knowledge-center/digital-knowledge/data-center
- Saudi National Data Center Strategy, June 2025, 1.5 GW by 2030 — https://www.trade.gov/market-intelligence/saudi-arabia-ict-new-data-center-strategy-accelerate-ai-and-cloud-expansion
- Asharq News via MOFCOM, 60 data centers and 3.4 GW planned capacity, 2026-06-04 — https://jedda.mofcom.gov.cn/jmxw/art/2026/art_31b51012a2254ad9898be03267203114.html
- Global Data Center Hub, LEAP 2026 announced total and AWS Saudi region investment — https://www.globaldatacenterhub.com/p/saudi-arabias-leap-2026-15-billion
- World Bank, World Development Report 2026 on artificial intelligence (via MOFCOM) — https://sa.mofcom.gov.cn/ztdy/art/2026/art_946f064f186d4d0bb2e5743bbec6d3f4.html
- IEA, Powering Artificial Intelligence, 2026-10-01 — https://www.iea.org/reports/powering-artificial-intelligence/executive-summary
- CBRE, Global Data Center Trends 2026 — https://www.cbre.com/insights/reports/global-data-center-trends-2026
- Saudi Energy Consulting, HUMAIN 6.6 GW pipeline and grid constraint analysis — https://saudienergyconsulting.com/insights/articles/humains-6-6-gw-ai-campus-dream-can-saudi-arabia-power-the-data-center-rush
- Arab News, data center desk — https://www.arabnews.com/tags/data-center