The Cabinet Decree
In late 2025, the Saudi cabinet formally designated 2026 as the national Year of Artificial Intelligence. The decree is not symbolic. It commits every government ministry — finance, health, education, energy, defense, transportation — to specific AI deployment milestones during the calendar year. It signals to the Saudi private sector and international investors that AI infrastructure is now the central organizing project of Saudi statecraft, replacing the megaproject narrative that defined the early Vision 2030 period.
The mechanism matters as much as the message. National-year designations are a familiar instrument of Saudi statecraft, but this one carries binding operational content: ministry-level milestones create accountability that a strategy document cannot, and the decree elevates AI from one of many Vision 2030 sectors to the primary economic transformation vehicle of the current phase. For a state apparatus in which ministerial performance is measured against Vision 2030 scorecards, attaching AI deployment targets to every portfolio converts the buildout from a technology ministry’s project into whole-of-government machinery. It is the institutional expression of the same reweighting visible in the capital flows — PIF’s $77 billion Humain commitment, the $20 billion-plus hyperscaler stack, the LEAP announcement totals that now exceed FII’s deal volume.
The Inflection Point
The Year of AI 2026 sits at a deliberate inflection point. By the end of 2025, Humain was operational, the first NVIDIA Blackwell GPUs had shipped, the export-control regime had stabilized, and the National Data Lake covered 430+ government systems. The 2026 designation reframes the buildout from “planning and procurement” to “execution and deployment” — the question is no longer whether the infrastructure will exist, but whether it will run AI workloads at scale.
Each of those preconditions was hard-won during 2025. Humain launched in May at the US-Saudi Investment Forum with the NVIDIA partnership (18,000 GB300 systems initially, a 600,000-GPU pipeline over three years), the Google Cloud $10 billion Dammam hub, and the AWS $5.3 billion region announced in a single window. The export regime stabilized in November, when US Commerce approved 35,000 GB300 systems for Humain — the largest sovereign GPU procurement outside the US and China — alongside Saudi Arabia’s designation as a major non-NATO ally. Groq’s $1.5 billion inference facility with Aramco Digital went operational in December. The decree, in other words, was issued at the precise moment when every input the buildout needed — silicon, capital, regulation, alliance — had been secured, and the only remaining variable was execution.
The Infrastructure Milestones
The visible milestones for 2026 are concrete. Humain’s first data center campus in Riyadh is scheduled to go operational, beginning at 100 MW and ramping to 200 MW, loaded with the initial 18,000 NVIDIA GB300 systems. The Hexagon government data center (480 MW, world’s largest sovereign DC) is scheduled to come online in early 2026 — roughly four to five times the estimated scale of the largest comparable US government facilities. Microsoft’s Saudi cloud region is targeted for Q4 2026 with West and East availability zones. LEAP 5 in April 2026 drew 201,000+ visitors and 1,900 institutional investors, generating cumulative LEAP investment commitments above $42 billion since 2022.
The 2026 calendar extends well beyond those anchors. The AWS Saudi region, funded by the $5.3 billion commitment, reaches operational status in 2026 with the Humain AI Zone construct inside it. Qualcomm’s 200 MW inference deployment on AI200 and AI250 rack systems begins in 2026. The Humain-xAI joint venture’s 500 MW Riyadh campus carries a 2026 target year, as does Gulf Data Hub’s 200 MW Riyadh facility backed by KKR’s $2 billion partnership. The SDAIA sovereign AI factory — up to 5,000 Blackwell GPUs under government authority, separate from Humain’s commercial fleet — moves through deployment. Not every target will land inside the calendar year; the point of the density is that enough of them landing converts Saudi Arabia’s announced pipeline into measured, operational megawatts.
The Integration Milestones
Less visible — but more operationally important — are the integration milestones. SDAIA’s National Data Lake is being extended to cover additional ministries, and its multi-petabyte substrate is transitioning from distributed hosting to centralized hosting at Hexagon — one of the structural operational milestones of the year. The SAMAI talent program is targeting 100,000 AI specialists by 2030, with the 2026 milestone being expansion of the training pipeline: from the 11,000+ specialists SDAIA reported by 2024, reaching the 2030 target requires a cadence of roughly 15,000-18,000 net new specialists per year, layered on top of SAMAI’s 1 million-plus participants in broad AI literacy training and MCIT’s 300,000 technology worker target.
Allam, the Arabic-first foundation model, is being deployed across consumer (Humain Chat), enterprise, and government applications. The 34B-parameter model’s government track runs on Data Lake-derived training data under full SDAIA control — administrative document processing, citizen-services chatbots, ministry decision support — while the consumer track puts sovereign Arabic AI into daily use. The integration milestones are the ones that determine whether the infrastructure milestones matter: a 480 MW facility is a shell until the Data Lake runs on it, and a 200 MW campus is stranded capital until models, customers, and operators load it. Execution in 2026 is measured in workloads, not ribbon-cuttings.
The Model and Application Layer
The Year of AI is also the year the Saudi model stack goes into production breadth. Allam anchors the sovereign track, but 2026 layers frontier partnerships on top of it. The Humain-xAI joint venture includes a country-wide deployment of Grok across Saudi consumer and enterprise users through Humain’s distribution channels — the dual-model approach pairing Allam for Arabic-first and sovereignty-sensitive applications with Grok where frontier capability matters more than control. The Adobe-Qualcomm-Humain partnership, announced in November 2025, extends Allam into Adobe’s creative tooling — Photoshop, Illustrator, Premiere — putting Arabic-language AI features on Saudi-hosted Qualcomm hardware inside globally distributed commercial products.
The hyperscaler regions complete the model surface: Azure OpenAI Service carries GPT-series models into Saudi enterprise on compliant in-Kingdom infrastructure, AWS Bedrock carries Anthropic’s Claude, and Google Vertex carries Gemini. By the end of 2026, a Saudi ministry or enterprise can reach sovereign Arabic models, US frontier models, and specialty inference — Groq LPUs for high-throughput Arabic serving, SambaNova RDUs for SDAIA’s specialized training — without workloads leaving the Kingdom. No other market outside the US and China will offer that breadth on domestic soil, which is precisely the position the Year of AI was designed to consolidate.
The Demand Question
Infrastructure and models are the supply side; the Year of AI’s quieter test is demand. The enterprise vector is already visible in the market data — Saudi banks, the telecom operators (STC, Mobily, Zain), and industrial conglomerates like SABIC and Aramco are pulling AI capacity onshore for both latency and data-residency reasons, and the Saudi data center market’s projected growth from $1.33 billion in 2024 toward $3.9 billion by 2030 assumes that enterprise adoption compounds through the decade. Aramco’s industrial AI portfolio — seismic analysis, refinery optimization, predictive maintenance — supplies the flagship use cases, with the Groq-Aramco Digital inference facility serving EMEA and South Asia from its Saudi base.
Government demand is mandated by the decree itself; enterprise demand must be earned. The distinction matters for how 2026 gets scored. Ministry milestones will be met, in some form, because the cabinet requires it. The more informative signal is commercial: whether Humain’s capacity fills with paying tenants, whether GroqCloud and the Qualcomm inference layer attract regional customers, and whether Saudi enterprises convert Cloud First-style pilots into production AI spending. A Year of AI that produces mandated government deployments but thin commercial uptake would satisfy the decree while missing its point — the token-export economy the buildout exists to create.
LEAP 5: The Showcase
LEAP 5 in April 2026 functioned as the Year of AI’s public showcase. The 201,000+ visitor count represents a 50%-plus year-over-year increase; the 1,900 institutional investors mark LEAP’s maturation from technology conference into sovereign-AI investment forum; the 1,800 brand exhibitors included most major US, European, and Asian AI companies. The announcements extended the Humain partnership stack, deepened export-control alignment with US chip vendors, and opened a new category — AI-enabled industrial automation across manufacturing, logistics, and energy operations, where Saudi Arabia’s economic concentration creates natural market pull.
LEAP’s structural role in the Year of AI is coordination. Because major commitments are systematically calendared to surface at the conference, LEAP functions as the annual public audit of buildout velocity: the year-over-year announcement total is the single most legible indicator of whether momentum is compounding or stalling. For the Year of AI specifically, LEAP 5 was the moment the 2026 thesis faced the market — and the attendance, investor presence, and announcement volume all registered acceleration rather than fatigue.
Ministry Execution and the Government Stack
The decree’s ministry-level milestones ride on a government AI stack that matured through 2025. The Cloud First Policy, in force progressively since 2021, gives every ministry a mandated migration path. The Data Lake gives them a shared analytical substrate. Hexagon gives them sovereign compute at hyperscale, and the SDAIA AI factory gives them government-controlled training and inference capacity. On top of that substrate, the enterprise vendors are positioning for ministry deployments: IBM’s watsonx targets government and financial-services workloads, Databricks’ platform partnership with SDAIA serves government and enterprise data workloads, and Salesforce’s $500 million Hyperforce commitment carries Saudi data-residency compliance for citizen-facing services.
The two-tier architecture keeps the sovereignty boundaries clean. Sensitive workloads — the Data Lake, defense applications, citizen data — run on SDAIA-controlled infrastructure with Saudi personnel and no foreign provider in the loop. Commercial and productivity workloads run on hyperscaler regions under PDPL and KSA-RoD compliance. The Year of AI’s ministry milestones are, in effect, the first full-scale test of whether that architecture works under operational load rather than in design documents.
The Coordination Test
The Year of AI is also when the buildout’s three supply-side constraints — chips, power, and facilities — must stay synchronized at scale. GPU delivery cadence is bounded by NVIDIA’s production capacity, TSMC’s CoWoS packaging throughput, and HBM3e memory supply, even with the 35,000-system export approval in hand. Power delivery is bounded by Saudi Electricity Company interconnection timelines, which stretch beyond nine months for the 50-200 MW loadouts Blackwell-class deployments require. Facility commissioning is bounded by EPC execution, fiber provisioning from STC and Mobily, and cooling capacity.
Any one of the three slipping strands the other two: chips without energized facilities sit in warehouses, facilities without chips depreciate empty, and power contracted without either wastes sovereign capital. 2026 is the first year the coordination problem runs at full scale — 100-plus megawatts energizing per quarter across multiple operators — rather than in the forgiving sequencing of a planning phase. This, more than any single facility opening, is what “execution” means in the Year of AI: the demonstration that the Saudi state can synchronize procurement, energy, construction, and talent on hyperscaler cadence.
The Window
The Year of AI is also a competitive window. The UAE’s parallel buildout (G42, Stargate, Microsoft) hits its own deployment milestones in 2026. China’s Huawei Ascend is rolling out to second-tier markets. Europe’s AI Act compliance overhead is constraining hyperscaler capacity in the EU. If Saudi Arabia executes 2026 cleanly, it credibly establishes itself as the third pole of global AI compute. If it slips, the UAE absorbs the regional centerpiece role.
The window logic is asymmetric in Saudi Arabia’s favor on inputs and against it on time. On inputs, no competitor pairs the Kingdom’s energy economics ($20-50/MWh industrial power), sovereign capital, and export-approved silicon pipeline. On time, first-mover advantages in regional AI services compound: the jurisdiction that hosts the region’s inference capacity, developer ecosystem, and enterprise AI relationships first becomes the default, and defaults are sticky. The November 2025 export framework also carries an implicit probation — the approvals assume Saudi facilities deploy on schedule and under the agreed conditions. Clean 2026 execution strengthens the case for the next tranche of the 600,000-GPU pipeline; visible slippage invites both US policy second-guessing and UAE positioning.
The structural contrast with the UAE sharpens the stakes. The Emirati model distributes its buildout across G42, Microsoft, and the Stargate architecture — multiple operators, multiple failure domains, faster individual decisions. The Saudi model concentrates capital, infrastructure, energy, and regulation under a single sovereign apparatus, which compresses coordination costs but means a single execution year — this one — carries the reputational weight of the entire program. 2026 is when the concentrated model proves it can ship, or doesn’t.
What to Watch
The Year of AI resolves into a handful of measurable indicators. Megawatts energized: Humain Riyadh’s ramp from 100 to 200 MW and Hexagon’s 480 MW commissioning are the two anchor data points. Regions launched: AWS operational status and Microsoft’s Q4 2026 target. Workloads live: the Data Lake’s Hexagon migration, Allam’s deployment breadth across ministries, and the first commercial customers on the Qualcomm and xAI capacity. Talent cadence: SDAIA’s annual specialist count against the 15,000-18,000 per year requirement. And the LEAP 2027 announcement total, which will record whether the 2026 execution year built or burned investor confidence.
The Cabinet decree was the easy part. Execution across 12 months is the hard part — and for the first time in the Saudi AI buildout, the results will be measurable in operational megawatts, live workloads, and audited milestones rather than announcements. That shift from promise to proof is what the Year of AI was designed to force.