On 13 May 2025, Crown Prince Mohammed bin Salman launched HUMAIN — a full-stack artificial intelligence company owned by PIF, created from the merger of the Saudi Company for Artificial Intelligence (SCAI), SDAIA’s model development team, and elements of Aramco Digital. Eleven months later, the company has secured 211 plots of land across Saudi Arabia with access to 14 gigawatts of power capacity, signed $23 billion in technology agreements with NVIDIA, AMD, AWS, and Qualcomm, invested $3 billion in Elon Musk’s xAI, broken ground on twin 100 MW data centre campuses in Riyadh and Dammam, and announced plans for 6.6 GW of AI compute capacity by 2034 at an estimated total cost of $77 billion.
CEO Tareq Amin — named to TIME’s 100 Most Influential People in AI in 2025 — described the ambition with a sentence that captures both its scale and its audacity: “We will build in one year what Saudi Arabia built in twenty.” The company’s mission: to make Saudi Arabia the world’s third-largest AI provider, behind only the United States and China, processing 7 per cent of global AI training and inference by 2030.
The ambition is not hyperbole. It is a capital allocation decision backed by the world’s fourth-largest sovereign wealth fund, executed in a country with the lowest solar power costs on earth, and timed to a moment when global AI compute demand is growing at 34 per cent annually and exceeding supply in every major market.
The Infrastructure Machine
HUMAIN’s data centre pipeline is the largest single AI infrastructure programme outside the United States and China.
The near-term targets: twin 100 MW campuses in Riyadh and Dammam, expected to go live in Q2 2026. The intermediate target: 1.9 GW of AI compute capacity by 2030. The long-term target: 6.6 GW by 2034. The estimated total cost: approximately $77 billion — a figure that approaches NEOM’s total spending ($50 billion) and that, unlike NEOM, is directed at assets with identifiable customers and growing demand.
The land position — 211 plots secured across the Kingdom with access to 14 GW of power — provides the physical foundation. The financing includes a $1.2 billion framework from Saudi Arabia’s National Infrastructure Fund for 250 MW of capacity, with additional financing structures under development. HUMAIN is actively seeking US data centre equity partners to co-invest in the buildout.
The GPU deployment is the most commercially significant element. HUMAIN will deploy 600,000 NVIDIA GPUs — including the latest GB300 NVL72 platforms and Quantum-X800 InfiniBand networking — over three years across Saudi Arabia and the United States. The initial cluster of 18,000 NVIDIA GB300 GPUs will serve xAI as its first major customer, powering the 500 MW+ flagship facility that the HUMAIN-xAI partnership is developing.
The US Commerce Department authorised the export of advanced semiconductors to HUMAIN (and G42 in the UAE), with each approved for the equivalent of up to 35,000 NVIDIA Blackwell GB300 chips — estimated at $1 billion in value. The approvals, conditioned on rigorous security and reporting requirements, followed Crown Prince MBS’s visit to Washington during which the Kingdom pledged $1 trillion in US spending.
Power as the Differentiator
Saudi Arabia’s National Renewable Energy Programme has produced some of the lowest utility-scale solar tariffs in the world: Sudair came in at $0.01239 per kWh on a 25-year PPA; Al Shuaibah set a global record at $0.0104 per kWh in its 2021 tender; Sakaka began the cycle at $0.0234 per kWh in 2018 and now anchors the broader pipeline. An NVIDIA GB300 GPU draws 1.4 kW per chip; an NVL72 rack draws 135-140 kW. The economics that follow from those two numbers — Saudi power tariffs at a fraction of US or European industrial pricing, and AI workloads with electricity bills that dwarf every other operating expense — are the structural reason HUMAIN can credibly compete with US and Chinese hyperscalers despite starting from a smaller base.
The 211 land plots HUMAIN has secured across the Kingdom give it the physical footprint to convert that power-cost advantage into deployed capacity, and the stc Group joint venture for 250 MW of capacity (announced February 2026) is the first structural commitment to convert grid access into operating data centres.
The Partnership Stack
HUMAIN’s technology partnerships span every major AI infrastructure provider — a breadth that reflects both the company’s ambition and the commercial appeal of a $77 billion infrastructure programme to technology vendors.
NVIDIA: 600,000 GPUs, including GB300s and Quantum-X800 networking. The partnership, announced at the US-Saudi Investment Forum, positions HUMAIN as one of NVIDIA’s largest non-US customers. NVIDIA CEO Jensen Huang has publicly endorsed the Saudi AI strategy.
AMD + Cisco Joint Venture: $10 billion strategic collaboration announced November 2025. Build target: 1 GW of AI infrastructure by 2030, with the first 100 MW phase commencing operations in 2026 and AMD and Cisco as exclusive technology partners.
AirTrunk (Blackstone-backed): $3 billion partnership announced for state-of-the-art data centre construction in the Kingdom — bringing Blackstone’s global hyperscale build expertise to HUMAIN’s pipeline.
MIS: SAR 1.88 billion ($501 million) design-and-build contract awarded February 2026 — the largest disclosed single contract for a Saudi engineering services firm tied to the HUMAIN buildout.
AWS: $5 billion-plus joint investment for an “AI Zone” in Riyadh, deploying up to 150,000 AI accelerators (NVIDIA GB300s and AWS Trainium chips). AWS committed to training 100,000 Saudi citizens in cloud and AI, with a dedicated initiative to train 10,000 women.
Qualcomm: MoU for advanced AI data centres, with HUMAIN committed to deploying up to 200 MW of Qualcomm AI systems starting 2026.
stc Group: Joint venture with stc’s centre3 to develop up to 1 GW of AI data centre capacity.
Groq: $1.5 billion commitment for AI inference infrastructure, partnering with Aramco Digital. Data centre in Dammam running 19,000 Groq LPUs.
xAI: $3 billion investment in xAI’s Series E round (February 2026), making HUMAIN a significant minority shareholder. Holdings converted to SpaceX shares upon the xAI-SpaceX merger. Joint development of a 500 MW+ flagship facility with xAI as the first major customer.
Total signed technology agreements: $23 billion. The figure excludes the $10 billion HUMAIN Ventures fund planned for AI startup investments across the US, Europe, and Asia.
The xAI Bet
HUMAIN’s $3 billion commitment to xAI’s Series E round, announced in February 2026, is the most strategically significant single transaction the company has executed. The Series E raised $20 billion in total at a $230 billion valuation. The investment established HUMAIN as a minority shareholder in the AI lab Elon Musk founded after departing OpenAI’s board — the lab that trained the Grok family of models and that has since merged with SpaceX in a transaction valuing the combined entity at approximately $1.25 trillion ($1 trillion SpaceX, $250 billion xAI). HUMAIN’s xAI position converted into combined-entity equity at the merger’s exchange ratio of 0.1433 SpaceX shares per xAI share, or $75.46 in cash per xAI share.
The transaction works on two layers. The financial layer is the equity stake itself. The strategic layer is what the partnership commits xAI to do operationally: use HUMAIN’s data centre infrastructure as a primary compute provider, anchoring the 500 MW+ flagship facility (announced November 2025) with a customer whose workload volume can fill the capacity from day one. The pattern matters. Most sovereign-backed infrastructure investments struggle to find committed anchor demand. HUMAIN started its build with the demand already signed.
ALLaM and HUMAIN ONE
HUMAIN’s software layer — built on top of the infrastructure machine — centres on ALLaM, the Arabic-first large language model.
ALLaM 34B: 34 billion parameters trained on approximately 3 trillion tokens of mixed Arabic and English data, per the model’s published research paper (arXiv, July 2024). The model understands Classical Arabic and regional dialects (Saudi, Egyptian, Jordanian, Lebanese), and was developed by an internal team that HUMAIN has publicly emphasised includes a meaningful proportion of Saudi PhD researchers and a deliberate gender mix.
HUMAIN Chat: launched 25 August 2025 as a free Arabic-first conversational AI, available on web, iOS, and Android in Saudi Arabia.
HUMAIN ONE: an agentic AI operating system launched at FII9 in Riyadh on 28 October 2025. The platform automates back-office operations — HR, finance, procurement — using the ALLaM model and an agentic orchestration engine.
HUMAIN has announced plans for a marketplace for developers and businesses, with initial focus on government, tourism, and healthcare. The first operational ALLaM deployments are running through HUMAIN Chat and the early HUMAIN ONE enterprise pilots.
The Talent Crisis
HUMAIN’s ambition confronts a constraint that money cannot immediately solve: the 50 per cent hiring gap in AI-related roles identified by Minister of Human Resources Ahmed Al-Rajhi. The Kingdom’s AI talent pool — 11,000 specialists trained under SDAIA programmes, with the SAMAI initiative reaching 1 million-plus participants — is growing but insufficient for an infrastructure programme that requires thousands of engineers to design, build, operate, and programme 6.6 GW of data centres.
The government’s training programmes (SAMAI, the National Skills Platform, partnerships with KAUST, Carnegie Mellon’s Qatar campus, and MIT) target 10,000 trained AI professionals by 2030. HUMAIN’s own workforce is heavily international, with high salaries and low retention — the same challenge that every Gulf technology employer faces. The competition for AI engineering talent is global, and Saudi Arabia’s lifestyle proposition (despite entertainment liberalisation and zero income tax) competes against Silicon Valley, London, and Singapore for engineers who can choose where they work.
The talent gap is HUMAIN’s most significant risk. Data centres can be built with construction labour. They cannot be operated without AI engineers. The 600,000 GPUs are hardware. The models trained on them, the applications built on those models, and the commercial customers served by those applications all require human expertise that the Kingdom has not yet developed at scale.
The Chip Approval
The US Commerce Department’s authorisation of advanced semiconductor exports to HUMAIN — for the equivalent of 35,000 NVIDIA Blackwell GB300 chips, estimated at $1 billion in value — is the most consequential US-Saudi technology decision in years. The approval signals a US policy choice to permit Saudi Arabia to build AI infrastructure at scale, despite earlier resistance to advanced chip exports to Gulf states whose end-use guarantees were considered insufficient. The conditions attached — security, reporting, and end-use verification requirements — establish a template that other sovereign AI projects seeking similar authorisations will need to meet. The Trump administration’s broader approach, using AI infrastructure access as a tool of foreign policy alignment with Saudi Arabia and the UAE as the test cases, has implications that extend well beyond either country.
HUMAIN vs G42
HUMAIN and G42 (UAE) are the Gulf’s two AI infrastructure contenders, and their competition defines the region’s technology landscape.
Both received equivalent of 35,000 NVIDIA Blackwell GB300 chip export approvals from the US Commerce Department. G42 partnered with OpenAI, Oracle, NVIDIA, and SoftBank for Stargate UAE — a 1 GW facility — while HUMAIN pursues a multi-partner model across NVIDIA, AMD, AWS, Qualcomm, and xAI. Saudi Arabia’s planned data centre capacity (2,200 MW) exceeds the UAE’s (500 MW) by a factor of more than four, reflecting the scale difference between PIF’s $1.15 trillion balance sheet and G42’s more modest corporate resources.
The UAE has first-mover advantages: the world’s first AI Minister (2017), the world’s first AI-dedicated university (MBZUAI), and the Falcon and Jais Arabic language models that preceded ALLaM. But HUMAIN’s resource advantage — the GPU count, the land position, the power access, the sovereign backing — gives it a scale advantage that G42 cannot match without additional sovereign support.
The competition is ultimately productive for the region. Two GCC countries building AI infrastructure simultaneously creates a regional compute cluster that could attract global AI companies seeking alternatives to US and Chinese data centres. The security requirements attached to the US chip export approvals — Saudi Arabia and the UAE must maintain rigorous controls over access to advanced semiconductors — add a compliance layer that, paradoxically, may increase international confidence in the region’s AI governance.
The Customer Side
The commercial model that converts $77 billion of infrastructure into recurring revenue depends on customers, and HUMAIN’s customer development is the most strategically watched workstream. The xAI partnership provides anchor demand. AWS provides hyperscaler demand. The captive Saudi government base — ministries, security services, sovereign-wealth-fund operations, and the major state-owned enterprises being steered toward HUMAIN-hosted AI services through procurement preferences — provides a structurally protected demand floor. The strategically important and most contested segment is the AI-native customer base: startups and growth companies whose value depends on the AI applications they build. The $10 billion HUMAIN Ventures fund is partly designed to bootstrap this segment by investing in AI startups that commit to using HUMAIN infrastructure as a condition of investment — a vertical integration that gives HUMAIN both equity upside and demand pull-through.
The Verdict
HUMAIN is the most important company PIF has created. Not because it is the largest (NEOM has consumed more capital), the most profitable (it has no revenue yet), or the most established (it is eleven months old). Because it is the first PIF portfolio company whose product — AI compute — has demonstrably growing global demand, structural cost advantages (Saudi Arabia’s sub-$0.02/kWh electricity), and a commercial model (infrastructure-as-a-service) that generates recurring revenue rather than depending on property sales, tourism arrivals, or government fleet orders.
The $77 billion price tag is enormous. But unlike NEOM’s $50 billion — which produced an airport, some roads, and 2.4 kilometres of foundation trench — HUMAIN’s capital will be deployed against assets that have customers before they are built. HUMAIN’s initial data centre capacity is already sold out. The 600,000 GPU deployment has identified users. The xAI partnership has a named customer. The AWS AI Zone has a commercial structure.
The comparison to NEOM is inescapable and instructive. NEOM asked: can we build a city that the world will want to live in? HUMAIN asks: can we build infrastructure that the world will want to compute on? The first question remains unanswered. The second is being answered — by every technology company that has signed a deal with HUMAIN, by every GPU manufacturer that has shipped to Saudi Arabia, and by every AI application that needs more compute than the current global supply can provide.
The talent gap is real. The execution risk is real. The geopolitical risk — demonstrated by the 894 intercepted missiles — is real. But the demand for AI compute is also real, and it is growing at 34 per cent annually. HUMAIN’s bet is that Saudi Arabia can capture a share of that demand before the competition catches up. The $77 billion says PIF believes the bet is worth taking. The 600,000 GPUs say NVIDIA agrees. The $3 billion in xAI says Elon Musk agrees. The $5 billion-plus joint investment says AWS agrees. The chip export approval says the United States government agrees, conditionally, that Saudi Arabia is the right partner for this allocation of strategic technology.
What is no longer in question is whether HUMAIN is real. The 211 land plots, the $23 billion in signed agreements, the AirTrunk-Blackstone $3 billion construction partnership, the MIS SAR 1.88 billion design-and-build contract, the stc 250 MW joint venture, the 18,000 GPUs landing in the first cluster, and the construction sites visible from satellite imagery across Riyadh and Dammam confirm that the company has crossed the boundary between announcement and execution. Few PIF portfolio companies have done so. The remaining question is whether the demand curve HUMAIN is betting on holds — and whether Saudi Arabia can build the talent base to operate the infrastructure it has committed to deploy.
This analysis draws on HUMAIN corporate announcements; NVIDIA GB300 specifications via Tom’s Hardware and Introl; Saudi solar PPA pricing from NS Energy (Sakaka), PV Magazine (Al Shuaibah), and ACWA Power (Sudair); the AMD-HUMAIN $10 billion strategic collaboration announcement via AMD Investor Relations and Cisco Newsroom; the AWS-HUMAIN AI Zone announcement (May 2025); the Blackstone/AirTrunk-HUMAIN $3 billion partnership announcement; the MIS SAR 1.88 billion contract via Arab News (February 2026); the stc-HUMAIN 250 MW joint venture via AGBI; CNBC and xAI disclosures on HUMAIN’s $3 billion Series E commitment and the $20 billion Series E raise at $230 billion valuation; CNBC and The Motley Fool coverage of the xAI-SpaceX merger at a $1.25 trillion combined valuation with the 0.1433 share-conversion ratio; CNBC coverage of the US Commerce Department chip export authorisation (November 2025); Data Center Dynamics on HUMAIN’s 211-plot land position and the 500 MW+ xAI facility; the ALLaM technical paper on arXiv (July 2024); TechCrunch on HUMAIN Ventures’ $10 billion fund; and TIME’s 100 Most Influential People in AI 2025 listing of Tareq Amin. Vision2030.AI is editorially independent and is not affiliated with HUMAIN, PIF, NVIDIA, or any official Vision 2030 entity.