Saudi Arabia’s national AI strategy is the Kingdom’s 2030 plan for data, compute, Arabic foundation models, AI governance, and sovereign infrastructure. Led by the Saudi Data and Artificial Intelligence Authority (SDAIA), the National Strategy for Data and AI (NSDAI) sets the policy architecture, while HUMAIN, the PIF full-stack AI company launched in 2025, turns that strategy into data centres, model deployment, and commercial AI products. The 2026 Cabinet designation of a Year of Artificial Intelligence marks the shift from framework to execution.
SDAIA and the National Strategy
SDAIA operates as the Kingdom’s apex AI institution, with a mandate spanning national AI strategy, data policy, and operational AI deployment. The authority reports directly to the Prime Minister, reflecting the strategic priority accorded to AI within the government hierarchy. Its current footprint encompasses three operational arms: the National Data Management Office (NDMO), which oversees data governance and sharing frameworks across government entities; the National Centre for Artificial Intelligence (NCAI), which drives research, model development, and applied AI; and the National Information Centre (NIC), which manages national databases and the underlying infrastructure that supports cross-government analytics.
The National Strategy for Data and AI (NSDAI), first published in 2020 and updated through subsequent five-year plans, organises the Kingdom’s effort around six pillars: ambition, competencies, policies, investment, innovation, and ecosystem. Headline targets aim to train more than 20,000 specialists by 2030 and build a top-15 global AI ranking. SDAIA reports more than 11,000 specialists trained as of early 2026, alongside the SAMAI workforce literacy initiative, which has reached more than one million participants in its first wave and expanded into a second phase (SAMAI 2) covering 11 government ministries.
SDAIA’s policy authority was sharpened in November 2025 with the release of the AI Adoption Framework, a structured baseline that public sector entities must follow for AI procurement, deployment, and governance. The framework codifies five pillars - data governance, model accountability, transparency, human oversight, and risk management - and shifts SDAIA from advisory body to de facto AI regulator for the public sector. For private vendors selling into government, the framework effectively becomes a procurement filter: vendors that cannot demonstrate compliance lose access to the Kingdom’s largest single AI buyer.
The 2026 Year of Artificial Intelligence designation, approved by the Cabinet in March 2026, layers coordination on top of the existing pillars, uniting ministries around shared AI programmes and fast-tracking high-impact deployments against the NSDAI’s 2030 endpoints.
ALLAM and Saudi LLMs
The single most visible product of SDAIA’s research arm is ALLaM, a series of Arabic-first large language models built by NCAI. The family includes 7B, 13B, and 70B variants initialised from Llama-2 weights, plus a 7B model trained from scratch. SDAIA mobilised 16 public entities to build the underlying corpus, producing what it describes as the world’s largest Arabic training dataset at roughly 500 billion tokens.
ALLaM is distributed through major hyperscaler model catalogues - available globally via Microsoft’s Azure AI Foundry, and via IBM’s watsonx platform under a royalty-free SDAI licence coupled with the Llama 2 community licence. The model has topped the Arabic MMLU benchmark in its size class, a performance signal that matters strategically: Arabic is structurally hard for models trained predominantly on English web text, and a credible sovereign Arabic model materially raises the cost for foreign frontier labs to capture the Arabic-speaking market.
The strategic rationale extends beyond benchmarks. First, ALLaM signals data sovereignty - sensitive government, judicial, and healthcare workloads can run on a model whose weights, training data, and update cadence sit under domestic control. Second, it underpins a regional play across roughly 400 million Arabic speakers. Third, it provides leverage in negotiations with frontier model vendors: a credible sovereign alternative is the strongest discipline on hyperscaler pricing.
ALLaM’s commercial deployment now sits within HUMAIN, which SDAIA describes as the production and distribution channel for advanced Saudi AI models, while NCAI retains research authority. That handoff splits Saudi Arabia’s LLM stack into a research arm (SDAIA/NCAI) and a commercial arm (HUMAIN) - structurally similar to how China separates state institutes from commercial champions.
HUMAIN and AI Investment
HUMAIN is the Public Investment Fund’s full-stack AI national champion, launched by Crown Prince Mohammed bin Salman in May 2025 and chaired directly by him. The company’s remit is unusually broad for a single firm: data centre construction, cloud infrastructure, foundation model development, and applied AI products and services. CEO Tareq Amin has stated publicly that HUMAIN’s ambition is to be the world’s third-largest AI provider behind the United States and China.
The capital stack behind that ambition is substantial. PIF and Saudi Aramco signed a non-binding term sheet in 2025 for Aramco to take a significant minority stake in HUMAIN, with PIF retaining majority control and both shareholders contributing AI assets, capabilities, and talent. HUMAIN has announced approximately $23 billion in strategic technology partnerships and a separate $10 billion venture fund for AI startups. A $1.2 billion funding round was confirmed during 2025 to expand AI infrastructure, and Blackstone’s AirTrunk signed a $3 billion data centre joint venture with HUMAIN in October 2025.
These figures sit within the broader $40 billion AI fund first reported in 2024, where PIF was in early talks with Andreessen Horowitz to co-anchor the largest single AI fund globally. The cumulative pipeline of disclosed Saudi AI commitments - HUMAIN’s announced spend, Aramco’s data centre capex, the PIF/a16z fund, plus hyperscaler partnership deals - clears the $100 billion threshold over a multi-year horizon. That dollar volume reframes the strategy: Saudi Arabia is no longer a customer of global AI infrastructure but a potential price-setter.
The rollout is not without execution friction. Semafor reported in August 2025 that HUMAIN had already sold out the entire capacity of its existing and under-construction data centres - evidence of demand but also of the supply-side bottleneck that the strategy will live or die on. The capacity targets are staged: 1.9 gigawatts by 2030, scaling to 6 gigawatts by 2034, which would represent roughly 6% of projected global AI compute supply at that horizon.
Data Centres
The compute build-out is the most physically visible dimension of the AI strategy. HUMAIN has begun construction on two large campuses comprising 11 data centres, each rated at 200 megawatts, with first facilities in Riyadh and Dammam targeted for 2026 operation. The Hexagon facility, billed as the world’s largest government data centre at 480 megawatts, anchors the public sector workload.
NEOM has been redesignated, in part, as a data centre hub. DataVolt signed a $5 billion agreement with NEOM for a 1.5-gigawatt AI factory in the Oxagon industrial zone, with operations beginning in 2028. Aramco has committed to a $5 billion net-zero AI data centre at NEOM and a $3 billion Humain-AirTrunk-Blackstone campus, part of approximately $52-58 billion in Aramco capex earmarked for 2025. NEOM accounts for roughly 50% of upcoming national power capacity for data centres.
Aramco Digital has converted the company’s structural energy advantage into an AI infrastructure thesis. At LEAP 2025, the conference surpassed $20 billion in technology investment commitments, including a $1.5 billion deal between Groq and Aramco Digital to establish what could become the world’s largest AI inference data centre. The strategic logic is straightforward: cheap stranded gas plus desert real estate plus subsidised industrial power equals a structurally lower cost-per-token than US or European alternatives.
Industry research estimates that more than $6 billion in new data centre investment will flow into the Kingdom by 2027, taking upcoming capacity to roughly 2.7 gigawatts. That trajectory makes Saudi Arabia, alongside the UAE, the largest non-US, non-China AI compute build of the decade.
Talent and Education
Human capital is the binding constraint. Compute can be procured and models licensed; trained AI engineers and PhD-level researchers cannot be procured at scale. The talent strategy operates on three tiers.
Tier one is upstream university capacity. King Abdullah University of Science and Technology (KAUST), King Fahd University of Petroleum and Minerals, and King Saud University all run dedicated AI faculties. KAUST Academy delivers a multi-stage AI specialisation programme to undergraduates and recent graduates, layered onto existing computer science cohorts.
Tier two is professional reskilling. The Tuwaiq Academy partnership with Stanford embeds frontier curriculum into a domestic bootcamp model, supplemented by Alibaba Cloud-powered training labs at Saudi universities and an STC-backed delivery channel. SDAIA’s own AI and Data Bootcamps and the Saudi Electronic University’s Thakaa Camp round out a bootcamp ecosystem that independent reviews report at 85% job-placement rates within six months. The SAMAI initiative provides the literacy floor - foundational AI fluency for civil servants and the general workforce.
Tier three is international recruitment. Premium residency, internationally competitive compensation, and the institutional credibility of KAUST and SDAIA have produced measurable inflows of foreign AI researchers, though precise hiring numbers are not consistently disclosed. Scholarship programmes channel Saudi students into top US and European AI programmes with return obligations, building a long-cycle pipeline that will mature in the late 2020s.
The honest assessment is that the talent gap remains large. SDAIA’s 20,000-specialist target by 2030 is demanding, and competition for senior AI engineers from US frontier labs, Chinese national champions, and the UAE’s G42 has compressed margins on hiring. The Kingdom’s structural advantages - tax-free packages, scale of work, and direct access to sovereign capital - are real but not infinite.
International Partnerships (NVIDIA, Microsoft)
The Kingdom’s partnership architecture follows a deliberate hub-and-spoke pattern: anchor agreements with the US technology stack, supplemented by selective European and Asian relationships, all routed through SDAIA, HUMAIN, or PIF rather than ministry-by-ministry procurement.
NVIDIA is the most significant single relationship. The May 2025 Saudi-US Investment Forum produced a framework for HUMAIN to deploy up to 600,000 NVIDIA AI accelerators over three years. The first concrete tranche was 18,000 GB300 Blackwell chips, with the December 2025 first shipment confirmed by AGBI. The US Commerce Department’s November 2025 authorisation of 35,000 Blackwell-class chips each to G42 and HUMAIN - in exchange for Gulf commitments to limit Chinese technology integration - moved the partnership from commercial to geopolitical. Jensen Huang named HUMAIN three times on NVIDIA’s November 2025 earnings call, confirming the relationship’s strategic weight.
Microsoft’s commitments are concentrated on cloud and applied AI distribution rather than chip supply. Public reporting cites an estimated $2.2 billion five-year Microsoft commitment, and ALLaM’s deployment on Azure AI Foundry effectively makes Microsoft the global distribution channel for Saudi sovereign AI. Google Cloud, Oracle, and AWS have each landed significant infrastructure deals.
Cross-stack diversification is deliberate. AMD has been added as a HUMAIN compute partner to reduce single-vendor risk. Groq’s inference-specialist hardware is being deployed via Aramco Digital. The pattern resembles a sovereign procurement strategy more than a commercial cloud roadmap - it optimises for capability redundancy and negotiating leverage rather than vendor consolidation.
AI in Vision 2030 KPIs
Vision 2030 does not include a single headline AI-share-of-GDP target on par with the tourism (10% of GDP) or private-sector (65% of GDP) KPIs. The AI contribution is instead distributed across multiple programmes: the National Strategy for Data and AI, the National Industrial Development and Logistics Programme, the Financial Sector Development Programme, and the Quality of Life Programme. SDAIA’s internal targets project AI to add a meaningful but not yet officially headline-pegged percentage to GDP by 2030.
The Vision 2030 2025 annual report indicates 93% of headline KPIs met their 2025 milestones, and AI-relevant indicators are clustered in digital infrastructure, technology contribution to non-oil GDP, and ICT employment categories. Saudi Arabia’s data centre footprint and AI-trained workforce metrics are now central to those bundles, which means underperformance in AI deployment mechanically threatens broader Vision 2030 dashboard outcomes.
A reasonable read is that AI is an enabling layer for nearly every Vision 2030 KPI rather than a standalone target line: it underpins the smart cities programme through NEOM and Riyadh, the healthcare quality bundles via SEHA Virtual Hospital and clinical decision tools, the productivity uplift assumed in the private sector GDP target, and the e-government transformation tracked by the Digital Government Authority.
Recent Developments 2024-2026
The pace and visibility of Saudi AI activity has accelerated sharply over the last 18 months. Key milestones include:
- March 2024: PIF and Andreessen Horowitz reported in early talks on a $40 billion AI fund, the largest single AI-focused fund globally.
- September 2024: ALLaM goes live on Microsoft Azure AI Foundry, providing global commercial distribution.
- May 2025: HUMAIN launched at the Saudi-US Investment Forum, alongside NVIDIA framework for up to 600,000 chips and Trump administration AI Zone announcement in Riyadh.
- August 2025: HUMAIN reports its existing data centre capacity is fully sold out; Aramco-PIF term sheet for HUMAIN minority stake announced.
- October 2025: Blackstone-AirTrunk signs $3 billion HUMAIN data centre joint venture.
- November 2025: SDAIA AI Adoption Framework published as mandatory baseline for public sector entities; US Commerce Department authorises 35,000 Blackwell-class chips each for HUMAIN and G42.
- December 2025: First NVIDIA GB300 shipment to HUMAIN confirmed; HUMAIN venture fund $10 billion commitment formalised.
- March 2026: Saudi Cabinet designates 2026 the Year of Artificial Intelligence; SAMAI 2 launched with 11 ministries.
- April 2026: Hexagon data centre (480 MW, world’s largest government facility) and Shaheen III supercomputer operational milestones reported.
The cumulative effect is that the Kingdom moved from policy-led AI strategy in 2019-2023 to capital-and-construction-led execution in 2024-2026. The risk-reward profile changes correspondingly: returns now hinge on delivery, not announcements.
Competition vs UAE
The competition with the UAE is the defining strategic context for Saudi AI policy and is best understood as parallel rather than zero-sum. Both countries have publicly committed in the range of $100 billion to AI projects. Both have flagship national champions - HUMAIN in Saudi Arabia, G42 in the UAE - and both are now confirmed as US-approved buyers of the most advanced NVIDIA Blackwell chips after the November 2025 Commerce Department authorisations.
The UAE’s Stargate UAE project, announced in 2025 with OpenAI, Oracle, NVIDIA, and Cisco, targets a 1-gigawatt AI compute cluster with the first 200-megawatt phase live in 2026 and a headline figure of $500 billion over the lifecycle. HUMAIN’s 1.9-gigawatt-by-2030 trajectory is broadly comparable in capacity but slower in initial phasing, reflecting Saudi Arabia’s larger physical build-out across multiple sites versus the UAE’s concentrated Abu Dhabi footprint.
The deeper differentiation is structural. The UAE under G42 has built a more open international compute play, anchoring partnerships with US frontier labs and multinational enterprise customers. Saudi Arabia’s HUMAIN combines an international commercial layer with explicit national-champion characteristics: Aramco-integrated energy supply, ALLaM-anchored Arabic sovereign LLM, and direct chair-level supervision by the Crown Prince. The UAE leans toward a Singapore-style global hub model; Saudi Arabia leans toward a Korea-style chaebol model with state coordination.
For investors and partners, the practical upshot is that the two markets are increasingly forced to specialise. Saudi Arabia’s structural advantages are scale of capital, energy cost, and Arabic-language data depth. The UAE’s are speed of regulatory execution, English-language enterprise market access, and earlier-stage cloud partnerships. Both will succeed in different segments; the rivalry compresses pricing in both directions.
Risks
The risk register is meaningful and deserves explicit treatment.
Execution risk is the largest. HUMAIN’s 1.9-gigawatt-by-2030 target requires roughly 200 megawatts per year of new capacity from a near-zero base, plus the human capital, supply chain, and energy interconnection to support it. Hyperscaler capex globally has run into transformer shortages, GPU supply constraints, and multi-year permitting delays. Saudi centralised decision-making mitigates some of this; the underlying physical bottlenecks do not disappear.
Geopolitical risk is structural. US chip export controls remain the single largest exogenous variable. The November 2025 authorisations for HUMAIN and G42 were conditional on Gulf commitments to limit Chinese technology integration. Any future US administration reversal, or any Saudi diplomatic posture toward China that triggers Section 1758 reviews, could constrain access to frontier accelerators with limited notice.
Concentration risk runs both directions. Saudi AI is concentrated in HUMAIN, and HUMAIN is concentrated on NVIDIA - if either fails to scale, the headline strategy falters. Conversely, NVIDIA and Microsoft are now meaningfully exposed to single-customer Saudi revenue lines.
Talent risk is persistent. Despite the SAMAI rollout and KAUST programmes, the Kingdom remains a net importer of senior AI talent. Compensation arbitrage works for now; if US frontier lab pay continues its current trajectory, the Saudi premium required to compete will rise.
Demand risk is underappreciated. HUMAIN’s reported sell-out of existing capacity reflects today’s demand environment, dominated by foundation model training and English-language inference. If the dominant workload shifts toward edge inference or smaller models, or if the global build-out enters a digestion phase, the Saudi build may face utilisation headwinds at the precise moment it is bringing the largest tranches online.
Regulatory and PDPL friction is a slow-burn risk. The PDPL is broadly aligned with GDPR but introduces local data residency and cross-border transfer restrictions that frictionally limit how international AI providers serve Saudi customers.
Outlook
Saudi Arabia’s AI trajectory positions the Kingdom as an increasingly significant player in the global AI landscape. The 2024-2026 acceleration has shifted the strategy from policy to execution, and the next phase will be defined by three measurable variables. First, capacity online: does HUMAIN convert its 1.9-gigawatt-by-2030 target into actually-energised, customer-loaded data centres on schedule, or does the schedule slip into the early 2030s? Second, model adoption: do ALLaM and its successors hold their lead on Arabic benchmarks and convert that lead into meaningful enterprise and government workload share against OpenAI, Anthropic, Google, and Chinese frontier labs? Third, talent stock: does the Kingdom hit its 20,000-specialist target by 2030, or does the gap widen as global AI compensation rises?
If these three lines hold, the Kingdom will likely emerge as the third-largest AI infrastructure provider globally by the mid-2030s, with disproportionate influence over Arabic-language and Islamic-world AI deployment. If any one slips materially, the strategy reverts to a regional AI hub with significant capital but a smaller global footprint than the headline numbers suggest. The risk-reward asymmetry, given the capital already deployed, now favours execution disclosure and benchmark publication rather than fresh announcements.
The Kingdom’s AI governance framework, Arabic-language capabilities, and research investment position it as a leader in AI development across the Arab world and broader Islamic community - aligned with Vision 2030’s broader objective of establishing Saudi Arabia as an influential global actor. At the current pace of execution, the probability of that outcome is meaningfully higher in 2026 than it was in 2024.
External references for further reading include SDAIA’s official site, NVIDIA’s announcement of the HUMAIN AI factory partnership, Bloomberg’s reporting on the Blackstone-HUMAIN data centre deal, and AGBI’s reporting on the $40 billion AI fund.