Saudi Arabia's HUMAIN — wholly owned by the $925B PIF and chaired by Crown Prince Mohammed bin Salman — placed an order for 18,000 Nvidia GB300 Grace Blackwell systems in late 2025 and operates ALLaM, the country's flagship Arabic LLM co-developed with SDAIA, under a Vision 2030 AI mandate that exceeds $100B in committed capex. NeuronFeed tracks 2 Riyadh-based AI companies, HUMAIN and Mozn, with sovereign vehicles rather than venture rounds dominating cap-table economics.

xAI signed a 500MW joint Riyadh facility with HUMAIN in November 2025, AWS committed up to 150,000 Nvidia GPUs to a dedicated Riyadh AI Zone announced at the US-Saudi Investment Forum, and Nvidia Nemotron now powers HUMAIN Chat aimed at the 400 million Arabic speakers Falcon also targets. Capital flows through PIF subsidiaries Sanabil ($3B+ deployed annually), Jada, and Alat ($100B mandate), with SVC anchoring the fund-of-funds layer and STV's $1B AUM leading early-stage rounds alongside Raed, Wa'ed, and Impact46. Google Cloud committed $10B in May 2025 to a Dammam region, and Groq announced a $1.5B Saudi data-centre allocation. The 2026 outlook is one of procurement at sovereign scale: HUMAIN's full GPU build-out spans several hundred thousand units across multiple kingdom sites by 2030.

Key trends in 2026

3 signals

HUMAIN's 18,000 GB300 order opens a 500MW Saudi AI factory pipeline

The late-2025 Nvidia tranche is the first of several scheduled across PIF-funded sites, with HUMAIN targeting 1.9GW of installed AI compute by 2030 — exceeding any single nation outside the US-China duopoly.

xAI's 500MW Riyadh joint venture and AWS's 150,000 GPU pledge formalize US-vendor concentration

Both deals announced at the November 2025 US-Saudi Investment Forum bind frontier-model supply to Saudi sovereign demand, with Groq's $1.5B and Google Cloud's $10B Dammam commitments adding 2026-2027 capacity.

ALLaM and the SDAIA stack monopolize Arabic-LLM deployment domestically

Co-developed with SDAIA and shipped via Nvidia Nemotron inside HUMAIN Chat, ALLaM is the mandated foundation model for Saudi public-sector AI procurement under Vision 2030's $100B+ digital line, locking out non-compliant alternatives through 2030.

FAQ

Frequently asked questions about Saudi Arabia

Why is HUMAIN treated as infrastructure, not a startup?

Because PIF capitalised it directly as a sovereign vehicle. HUMAIN owns data centres, the cloud layer, the ALLaM foundation model, and applications under one roof — a vertical stack that doesn't exist elsewhere in the GCC. Its 18,000-GB300 order and 500MW xAI facility are procurement decisions for the kingdom, not venture milestones.

How does Saudi compute compare to UAE Stargate?

Different shapes, similar scale. Stargate UAE is 1GW G42-built and OpenAI-operated. HUMAIN is targeting up to 500MW of its own AI factories plus xAI's 500MW joint build and AWS's 150,000-GPU Riyadh AI Zone. The UAE leans hyperscaler partnerships; Saudi runs more through HUMAIN as the sole national counterparty.

What does ALLaM bring that Falcon doesn't?

Saudi-specific data, SDAIA governance, and tighter integration with HUMAIN's product surface. Falcon is published as open weights from TII Abu Dhabi; ALLaM is being commercialised through HUMAIN Chat and enterprise deployments backed by Nvidia Nemotron training pipelines, with PDPL data-localisation requirements baked in.

Is there a venture layer outside HUMAIN and Mozn?

A thin one. STV, Raed, Wa'ed (Aramco-aligned), and Impact46 lead Riyadh seed and Series A rounds, with SVC anchoring the LP base and a16z and Sequoia increasingly co-investing later. Most capital nonetheless flows through PIF subsidiaries Sanabil, Jada, and Alat, so independent VC remains a fraction of the deployed total.