Wholly owned by the Public Investment Fund. Launched May 13, 2025. $77B infrastructure commitment. Target: third-largest AI provider behind only the United States and China.
Eleven data centers under construction across Riyadh and Dammam. 200 MW per facility. Cadence: +50 MW per quarter.
Humain at a Glance
Humain is the corporate vehicle through which Saudi Arabia intends to become the world’s third sovereign-AI power, behind only the United States and China. Incorporated in May 2025 as a wholly-owned subsidiary of the Public Investment Fund (PIF), it consolidates the Kingdom’s compute, model, and applied-AI ambitions inside a single entity that reports up to PIF Governor Yasir Al-Rumayyan and ultimately to Crown Prince Mohammed bin Salman, who sits as chairman. The company’s stated mandate is to deploy roughly $77 billion of capital expenditure across data centers, accelerator silicon, sovereign foundation models, and downstream applications by 2030, with an interim capacity target measured in single-digit gigawatts of IT load and high-six-figure GPU counts.
Where the Emirati G42 took the integrator path — assembling clouds, models, and chips through a portfolio of related companies — Humain has been engineered as a single P&L. Infrastructure, models, and applications all sit inside the same balance sheet. That structure is unusual globally and is the analytical key to the entity. It compresses the time between a sovereign capital decision and a deployed kilowatt of AI compute, and it gives the Saudi state a single counterparty for every hyperscaler, every chip vendor, and every foreign government conversation about Kingdom-resident AI.
Leadership and Governance
Humain’s chief executive is Tareq Amin, the former CEO of Rakuten Symphony and the architect of the Rakuten Mobile open-RAN buildout in Japan. Amin’s hiring in 2024 telegraphed the strategy. He is not a hyperscaler veteran; he is a greenfield-network operator whose career has been about deploying capital into physical infrastructure on compressed timelines. The board pairs him with PIF deputy governors, MCIT representation, and observer seats for the Saudi Data and AI Authority (SDAIA) — the institutional machine that wrote the 2020 National Strategy for Data and AI and that operates the National Information Center.
Reporting structure flows: Humain CEO -> PIF Governor Al-Rumayyan -> the PIF board chaired by MBS. SDAIA’s role is policy, standards, and the National Data Bank; Humain’s role is operating company. Communications Pro Authority (CST) and the Cloud Computing Special Economic Zone administered through the Economic Cities and Special Zones Authority (ECZA) provide the regulatory perimeter. The Ministry of Communications and Information Technology (MCIT) under Abdullah Alswaha handles cabinet-level coordination.
The $77 Billion Capex Envelope
The capex envelope is the most-cited figure in any Humain disclosure, but it conceals more than it reveals. Of the headline number, the largest individual buckets are data center civil and electrical works (roughly 35 to 40 percent), accelerator silicon procurement (roughly 30 to 35 percent), and software, models, and applied-AI workforce (the remainder). PIF is the equity anchor. The company is also expected to operate as a co-investor with hyperscalers — primarily AWS, Microsoft, and Google Cloud — that bring their own capex into Saudi-resident regions and availability zones. The on-balance-sheet number for Humain therefore understates the total compute deployed inside Saudi borders that the company will touch operationally.
The capital ladder is sequenced. The first tranche, broadly the 2025 to 2027 window, funds the initial hyperscale campuses, the first 35,000 NVIDIA GB300 systems disclosed under the May 2025 framework, and seed positions in AMD MI355X and Qualcomm AI200/AI250 inference fleets. The second tranche, 2027 to 2030, scales the campus footprint toward gigawatt-class loads, refreshes accelerator inventory to whatever the post-Blackwell generation looks like, and funds Humain’s foundation-model and applied-AI businesses to revenue. The third tranche, post-2030, is contingent on whether the export-control posture of the United States continues to permit Tier-2 access to frontier silicon and on whether the offtake market — sovereign-AI customers across MENA, South Asia, and Africa — materializes at the rates assumed in the original Vision 2030 thesis.
Strategic Partnerships
The partnership stack is the load-bearing structure of the company. NVIDIA is the anchor silicon supplier; the May 2025 framework allocates an initial 35,000 GB300 systems against a multi-year ceiling reported in some disclosures at up to 500,000 units, all subject to BIS license review under the AI Diffusion Tier-2 framework. AMD is the second source, with the MI355X line targeted at Humain’s training and HPC workloads. Qualcomm covers inference at scale through the AI200 and AI250 platforms, an unusually deep design-in for a company more associated with mobile silicon. Groq’s LPU architecture rounds out the inference layer for latency-sensitive Arabic-language workloads.
On the hyperscaler side, AWS announced its Saudi region with $5.3 billion of associated capex; Microsoft’s Azure footprint anchors on its $1.5 billion G42 investment that re-routes capacity into Humain-aligned KSA capacity; Google Cloud is committed to a Saudi region paired with AI-development support to Humain’s model team. Oracle is present through its OCI Riyadh region. xAI’s Saudi training-cluster footprint, deployed alongside Humain in 2025, gave the company a second pillar of frontier-model adjacency to complement the OpenAI-Microsoft axis that runs through Azure.
Underneath those marquee names sit the operational partners: Cisco for networking fabric, Schneider Electric and ABB for medium-voltage power and switchgear, Vertiv for thermal, ACWA Power and Saudi Electricity Company for the underlying generation and transmission, Aramco for hydrogen, gas, and the energy-systems backbone that ties the data centers to Saudi primary energy, and Saudi Telecom Company (STC) and its Center3 subsidiary for fiber and connectivity.
Capacity Targets and Site Map
Humain’s operational footprint is anchored by four campus programs. The Riyadh metro cluster, sited in proximity to the existing King Abdulaziz City for Science and Technology (KACST) complex and the SDAIA campus, is the headquarters compute footprint. The Eastern Province cluster, near Dammam, exploits the cheaper power and proximity to the Aramco gas grid. The NEOM-aligned program, executed in partnership with DataVolt and the NEOM Tech & Digital Company, is the long-duration moonshot — a 1.5 GW DataVolt anchor with optionality on additional Humain-direct loads. Hexagon, the formerly-Aramco-aligned data center program, is increasingly entangled with Humain’s pipeline and is reported at roughly 480 MW.
In aggregate, Saudi-resident IT load disclosed across announcements totals more than 6.6 GW of capacity at various stages of permitting, design, construction, and operation. Humain itself targets roughly 1.9 GW within its own consolidated footprint over the medium term, with GPU counts that the company has guided toward the 600,000-unit zone — a number that, if achieved, places Humain within the same order of magnitude as a single Microsoft Azure or AWS region’s accelerator fleet.
The Sovereign-AI Thesis
The strategic logic is straightforward to state and harder to execute. Saudi Arabia possesses three structurally rare inputs to sovereign AI at scale: very low marginal-cost power (gas, solar, and increasingly grid-connected nuclear), sovereign capital with a multi-decade tolerance for negative free cash flow, and political access to US frontier silicon that most other sovereigns cannot replicate. Humain is the corporate expression of the thesis that those three inputs, vertically integrated under a single P&L and pointed at the global Arabic-speaking market plus offtake customers across the Global South, can produce a third pole of frontier AI alongside the United States and China.
The thesis does not require Humain to outcompete OpenAI or Anthropic on frontier model quality. It requires Humain to be the lowest-cost provider of large-scale, sovereignty-respecting compute and Arabic-first applications to a customer set that is structurally underserved by US and Chinese clouds. The bet is that, over a 2026 to 2032 window, that customer set is large enough to absorb gigawatts of Saudi-resident capacity at margins that justify the capex envelope.
Risks
The risks fall into five categories. Export-control risk is the most material: the AI Diffusion framework’s Tier-2 status for Saudi Arabia is a political artifact, and a future US administration could narrow license availability, cap end-use, or impose verification regimes that slow shipment cadence. Execution risk is the second: Humain is being asked to execute a multi-year, multi-gigawatt, vertically integrated buildout against a labor pool that the Kingdom does not yet possess at scale. Demand risk is the third: the assumed offtake from MENA, South Asia, and African sovereign-AI customers is real but not yet contracted at the scale the capex envelope implies. Power and water constraints are the fourth: gigawatt-class data centers in 50C peak ambient conditions require thermal architectures and water budgets that are still being validated. Geopolitical realignment is the fifth: any breakdown in the US-Saudi strategic alignment that underpins the Major Non-NATO Ally framework would propagate immediately into the silicon pipeline.
What to Watch in 2026
The leading indicators for the next twelve months are licensing throughput at BIS, the operational cutover of the first GB300-anchored Humain campus, the disclosure of named anchor customers outside the Saudi state, the trajectory of Humain’s foundation-model program, and the ratio of announced-to-installed capacity across the campus pipeline. Each of those is tracked on saudicompute.com on a per-deal, per-facility, per-license basis.
Humain is the entity through which the Saudi compute thesis succeeds or fails. The other 130-odd entities tracked on this platform are partners, suppliers, customers, regulators, or competitors of Humain. Read this section as the operating company; read the others as the ecosystem around it.
The Foundation Model Program
Humain’s foundation-model program is the most-watched applied-AI workstream inside the company. The thesis is that an Arabic-first foundation model — trained on the largest curated Arabic corpus assembled to date, evaluated against MENA-region task benchmarks, and tuned for the dialectal diversity of Levantine, Egyptian, Gulf, and Maghrebi Arabic — has a structural addressable market that US and Chinese frontier-model programs underserve. The Kingdom’s bet is that, paired with the compute footprint Humain is building, the foundation-model program is the substrate on which the broader applied-AI business runs.
The program’s technical architecture pairs a frontier-class transformer backbone with a specialized data-curation pipeline run by SDAIA’s National Information Center and Humain’s own model-engineering team. The training infrastructure is the GB300-anchored Humain campus footprint, with AMD MI355X providing structural second-source training capacity. The evaluation harness combines public benchmarks (MMLU, BBH, MATH, HumanEval) with Arabic-specific evaluations and MENA-domain task suites covering banking, healthcare, government services, education, and energy. The deployment pattern is to operate the model behind Saudi-resident inference fleets — primarily Qualcomm AI200/AI250 and Groq LPU — for latency-sensitive Arabic-language workloads.
The competitive context is the UAE Falcon program (TII), the Jais model program (G42), the broader Arabic-language model ecosystem, and the dialectal-Arabic capabilities of US frontier models (GPT-class, Claude-class, Gemini-class). Humain’s competitive edge is not raw frontier-model quality — at least not initially — but the integration of Arabic-language capability with Saudi-resident inference economics and the broader applied-AI go-to-market that Humain controls.
Applied AI: The Sector Roadmap
The applied-AI layer of Humain is the commercial expression of the compute thesis. The company’s announced applied-AI roadmap covers six primary sectors: government services, banking and financial services, healthcare, education, energy, and entertainment-and-media. Each sector has a Humain product line, an internal P&L, and an external partnership architecture that pairs Humain with the relevant Saudi sovereign-corporate counterparty.
In government services, Humain partners with SDAIA, the National Information Center, and the broader e-government apparatus to deliver AI-augmented citizen services, government-internal productivity tooling, and the Arabic-first natural-language interfaces that underpin Vision 2030’s e-government targets. In banking, the partnership architecture pairs Humain with the Saudi Central Bank’s regulated-cloud envelope and the major Saudi banks (SNB, Al Rajhi, Riyad Bank). In healthcare, Humain partners with the Ministry of Health, the major Saudi healthcare networks, and the Saudi Health Information Exchange to deliver clinical-workflow AI inside the MoH-administered data-residency envelope.
In education, Humain partners with the Ministry of Education, the Tuwaiq Academy program, and the major Saudi universities (KAUST, KFUPM, KAU, KSU) to deliver AI-augmented learning, faculty productivity tooling, and the Arabic-first educational-content layer. In energy, Humain partners with Aramco, ACWA Power, and SABIC to deliver industrial-AI applications across upstream operations, downstream chemicals, and grid optimization. In entertainment and media, Humain partners with the Savvy Games Group and the broader Vision 2030 entertainment portfolio.
The applied-AI revenue trajectory is the most important commercial-traction metric for the company. The infrastructure layer carries depreciation and operating costs that the model and applied-AI revenue has to cover for the company to reach P&L sustainability inside the 2030 horizon. The platform tracks applied-AI customer disclosures and revenue-traction signals as they surface.
The Venture and Ecosystem Layer
Humain operates a venture and ecosystem layer that complements the operating-company P&L. The Humain Ventures arm invests in Saudi-domiciled AI startups, regional MENA AI companies, and selectively in global AI infrastructure plays where the strategic alignment with Humain’s core thesis is high. The fund’s deployment cadence is multi-hundreds-of-millions per year against a multi-billion-dollar ten-year envelope.
The ecosystem layer also includes the Humain partnership program, which structures relationships with hyperscalers, silicon vendors, ISVs, and applied-AI customers across the universe. The program’s structural function is to give external counterparties a single point of engagement with the Saudi compute economy, so that the multi-vendor, multi-partner architecture is operationally tractable for foreign companies that would otherwise face high coordination costs across the Saudi institutional landscape.
The Tuwaiq Academy partnership, the SDAIA AI Academy collaboration, and the broader skills-development workstream are paired with the venture and ecosystem layer. The thesis is that the Saudi AI labor pool has to scale at roughly the rate of the compute footprint, and that the venture, ecosystem, and skills-development workstreams together drive that scaling.
For deeper reading: