AWS and HUMAIN expand Saudi AI zone to as much as 50 MW

A dedicated AI zone, not just a general cloud region AWS and HUMAIN announced an expansion of their dedicated Saudi AI zone to as much as 50 megawatts of capacity, according to my.gov.sa's official summary of LEAP 2026's announcements , with a target tied to 2028. The announcement is distinct from AWS's separate general-purpose Saudi cloud region, covered in TechScoop's piece on the region's December 2026 launch : this AI zone is purpose-built for AI workloads specifically, rather than the broader mix of enterprise computing a standard cloud region supports. What sits inside the zone The expanded zone is built around three named components: Amazon's Bedrock platform, which gives developers access to a range of foundation models through a single managed service; ALLAM, an Arabic-focused AI model; and HUMAIN Fabric, HUMAIN's own compute infrastructure layer. Combining a general-purpose model-access platform, a language-specific model and a dedicated national compute layer inside one facility gives the zone a full-stack character rather than functioning as raw, undifferentiated compute capacity alone. Why megawatts is the number that matters here Capacity figures for AI infrastructure are typically expressed in megawatts rather than dollars because power draw, not server count, is what ultimately caps how much AI compute a facility can run. Modern AI training and inference hardware, particularly the GPUs at the centre of most large model workloads, draws substantially more power per rack than the general-purpose servers that filled data centres a decade ago, which means the megawatt figure attached to an AI zone is a more meaningful proxy for its real capacity than a simple server count would be. The same metric appears in the other data-centre announcements from LEAP 2026, including MIS's 192-megawatt plan and NHC Innovation's roughly 65-megawatt Khuzam project. Why HUMAIN is central to this specific deal HUMAIN is Saudi Arabia's state-backed AI company, and it appears as the domestic partner in more than one of LEAP 2026's headline deals, including Adobe's expanded Saudi partnership. TechScoop has separately covered HUMAIN's strategic investment in the Saudi AI company MOZN . Inside the AWS AI zone, the division of roles is straightforward: AWS brings Bedrock and its global platform, while HUMAIN brings HUMAIN Fabric, the domestic compute layer, with the resulting capacity located inside the Kingdom. A 2028 target, not an immediate one The target tied to this expansion sits around 2028, underscoring that this is a multi-year infrastructure build rather than an immediate capacity increase. That timeline is broadly consistent with how long large-scale data-centre and power infrastructure projects typically take to plan, permit, construct and bring online, and it places the AI zone's full 50-megawatt capacity several years out from LEAP 2026's announcement itself — a detail worth keeping in mind against any expectation that AI compute capacity of this scale appears overnight. For comparison, AWS's general-purpose Saudi cloud region is due to launch in December 2026 and Microsoft's Azure region in November, so the AI zone's full 50-megawatt capacity is scheduled to arrive after both general-purpose regions are live. Part of a bigger power and cooling story Every additional megawatt of AI compute capacity announced in Saudi Arabia — this AI zone included — comes with a proportional power and cooling requirement behind it. TechScoop connects that dynamic across this project and several others, including data-centre projects from MIS and NHC Innovation and cooling innovations like Ferveret's immersion-cooling technology, in a broader piece on how Saudi Arabia's AI boom is becoming an electricity and cooling story . What this means for developers and enterprises For a developer or enterprise team, having Bedrock available inside a dedicated Saudi AI zone means access to a managed way of building on top of multiple foundation models without needing to run and maintain that infrastructure independently, while still keeping the underlying compute and data inside the Kingdom. Pairing that general-purpose model access with an Arabic-specific model like ALLAM gives Saudi developers a genuine choice between a broad, general-purpose model layer and one purpose-built for Arabic-language use cases, depending on what a given application actually needs. How this fits the rest of AWS's Saudi build-out Seen alongside AWS's separate general-purpose cloud region, this AI zone expansion suggests AWS is deliberately building two distinct kinds of Saudi infrastructure at once: broad enterprise cloud capacity for general workloads, and a dedicated, AI-optimised zone for the compute-intensive training and inference work that Saudi enterprises and government bodies may need over the coming years. That dual build-out mirrors the approach hyperscalers have increasingly taken globally, where AI-specific capacity is treated as its own infras