Saudi Arabia's LEAP Data-Centre Buildout: The Megawatts Matter More Than the $15 Billion Headline

LEAP 2026 closed in Riyadh with nearly $15 billion of technology investments, agreements and announcements, according to Saudi Press Agency. The total spans AI infrastructure, data centres, cloud, technology manufacturing, venture capital and other programmes, so it should not be read as a $15 billion data-centre spending figure.

For Saudi Arabia's compute buildout, the more useful evidence is at project level. Several announcements now attach actual megawatt targets, launch dates or development stages to the broader ambition. Those numbers make it possible to distinguish infrastructure that is being built from cloud regions that are about to open and capacity that remains a longer-term target.

MIS put the clearest new capacity number on the table

Al Moammar Information Systems announced a $1.2 billion programme to expand data centres in Saudi Arabia and increase its operational capacity to a total of 192MW. The wording is important: 192MW is a target for total operational capacity, not evidence that 192MW is operating today.

That still makes the MIS announcement unusually useful. Many Gulf technology announcements disclose capital without power capacity, or power capacity without a delivery timetable. A target tied to an operator's total operational footprint can be tracked over time against commissioned facilities rather than treated as complete on announcement day.

NHC's 65MW project exposes a source-quality problem worth showing readers

NHC Innovation announced data centres at Khuzam Digital Valley with scalable capacity targeting 65MW by 2033. SPA's 1 September release put the project value at $880 million. Its 4 September closing summary described the same 65MW project at $800 million.

Gulf Business Review is not choosing one figure and silently discarding the other. Until NHC or another primary project document resolves the discrepancy, the capacity and 2033 target are more reliable reference points than a single dollar total. The conflicting values are also a useful reminder that event-roundup numbers should be checked against project-level releases before they enter permanent datasets.

AWS adds a launch date, while HUMAIN adds a separate AI-zone target

SPA says AWS's Saudi cloud region is due to launch in December 2026 as part of planned investment exceeding $5.3 billion. That should be tracked as a cloud-region availability milestone, not converted into megawatts because AWS has not disclosed a site-level MW figure in the cited material.

A separate AWS-HUMAIN plan targets up to 50MW of infrastructure within the first AI zone by 2028. These figures describe different things. The 50MW AI-zone capacity cannot simply be added to an undisclosed AWS region figure, and neither should be merged with HUMAIN's separate partnerships with Nvidia or DataVolt.

center3's 250MW is another programme that needs stage discipline

Saudi telecom infrastructure company center3 said it is developing AI-ready data-centre capacity starting at 250MW, alongside a partnership with HUMAIN spanning data centres, connectivity and go-to-market activity. The number is material, but the phrase 'starting at' is not a commissioning statement.

GBR therefore treats it as a development programme rather than 250MW of operating capacity. The same standard applies to every operator in the tracker. Announced design scale, under-development capacity and live capacity are separate states, even when all three appear in the same conference week.

The Red Sea buildout is moving one layer closer to delivery

Away from Riyadh, HUMAIN and DataVolt have already said 100MW is under development at Oxagon within a 360MW first phase. That is a stronger delivery signal because it identifies a site and a capacity currently being developed, although it is still not operating capacity.

Taken together, LEAP and Oxagon show why the Saudi AI infrastructure story is becoming more measurable. The market now has a mix of cloud launch dates, project MW, first-phase construction and long-term targets. The next editorial task is not to add all those numbers together, but to keep them separated long enough to see which ones become commissioned infrastructure.

What the $15 billion headline does and does not tell us

The conference total is useful as a measure of dealmaking intensity, but weak as an infrastructure metric. It mixes different sectors, transaction types and stages. One agreement can describe future spend, another a commercial partnership, another a plant and another a service launch.

For investors, customers and suppliers, the project table is more useful than the conference total. Power capacity, construction stage, operator, location and launch date determine whether an announcement becomes an investable supply chain, usable cloud capacity or simply another line in a national technology strategy.