Key points
- Nvidia says sovereign AI revenue more than tripled to over $30 billion in fiscal 2026. It grew another 80% year over year in the following quarter, with Nvidia infrastructure deployed across nearly 40 countries.
- Sovereign, hyperscaler, and neocloud demand overlap. The same dollar of spending often shows up in more than one company's numbers.
- Oracle's (ORCL) backlog reached $638 billion, but it includes overlapping demand from AI labs, hyperscalers, and infrastructure projects.
- Nvidia's original $100 billion OpenAI plan never became a firm commitment; OpenAI later announced a $30 billion Nvidia investment in its February 2026 funding round.
The AI buildout is usually told as a story about four American companies: Microsoft, Amazon, Alphabet and Meta. They are the biggest spenders, but they are not the whole market. Countries have started treating AI compute as strategic infrastructure, something to own rather than rent from a foreign cloud.
Nvidia (NVDA) puts a number on it. On its fiscal 2026 earnings call, finance chief Colette Kress said sovereign AI revenue more than tripled year over year to over $30 billion, close to 14% of the company's total, led primarily by customers in Canada, France, the Netherlands, Singapore and the UK. Projects underway in Saudi Arabia, the UAE and India sit in the pipeline behind that figure and could expand the category further, but Nvidia did not attribute the disclosed $30 billion to them. On its following call in May 2026, Nvidia said sovereign revenue grew more than 80% year over year, with infrastructure now deployed in nearly 40 countries.
The sovereign deals
| Deal | What it is | Status |
|---|---|---|
| Saudi Arabia, HUMAIN | State AI company (PIF); a first 18,000-chip Nvidia GB300 supercomputer, part of a five-year plan for up to 500 megawatts | Announced May 2025 |
| Stargate UAE | A 1-gigawatt Nvidia-powered cluster (G42, OpenAI, Oracle, Nvidia, SoftBank, Cisco) inside a broader 5-gigawatt UAE-US AI campus | First 200 MW phase due 2026 |
| US-Gulf chip framework | Commerce authorized G42 and Humain for up to 35,000 Blackwell-class chips each | Confirmed Nov 2025, under security conditions |
| France | €109B in announced French and foreign private investment into AI in France, led by a €50B UAE commitment | Announced targets, not booked revenue |
| EU AI Gigafactories | A July 2026 tender for up to seven facilities, backed by as much as €10B in public funding and expected to unlock more than €20B privately, part of the EU's broader €200B InvestAI ambition | Bids due November 12, 2026; selection expected in early 2027 |
| India | IndiaAI Mission (over $1B); an L&T gigawatt-scale sovereign AI factory | IndiaAI Mission funded; L&T venture proposed February 2026 |
These figures represent different levels of certainty. Some refer to firm chip orders carrying stated dollar values. The European figures are government mobilization goals rather than executed contracts. Headlines may place them side by side, but an order and an investment target should not be treated as equivalent.
The stocks it runs through
Much of this activity still passes through American technology companies. A single sovereign project might combine Oracle infrastructure, Nvidia chips and OpenAI software. CoreWeave or IREN might then supply capacity under a contract from Microsoft or another hyperscaler. Each company occupies a different link in the chain, so the reported figures cannot automatically be added to hyperscaler budgets as incremental demand. They show where projects are financed, contracted and delivered.
Nvidia is positioned to supply several of these projects. Oracle (ORCL) has become an AI cloud landlord almost overnight. Its remaining performance obligations, meaning contracted revenue the company has yet to deliver, stood at $638 billion in the quarter that ended in May 2026. That was a 363% increase, with large AI agreements accounting for most of the recent growth.
Then there are the neoclouds. They specialize in building and renting out large clusters of AI accelerators, and their growth can outrun diversified cloud companies, but their financing needs and customer concentration tend to run much higher too. CoreWeave (CRWV) reported $2.6 billion in revenue for the June 2026 quarter, up 112%, with a backlog near $104 billion. Nebius (NBIS) grew revenue 454% to $582 million and reaffirmed full-year guidance of $3.0 to $3.4 billion. IREN (IREN) started as a bitcoin miner and pivoted, signing a $9.7 billion, five-year cloud deal with Microsoft. We've laid out how these businesses work in our neocloud explainer.
What could go wrong
Neoclouds are especially exposed to customer concentration. A large portion of CoreWeave's backlog depends on Microsoft and OpenAI, leaving the company reliant on a small group of counterparties participating in the same spending cycle.
The second is circular financing, and the terms of these deals are narrower than the headline numbers suggest. Nvidia and OpenAI originally announced a nonbinding plan under which Nvidia could invest as much as $100 billion as 10 gigawatts of infrastructure were deployed. That arrangement never became a firm $100 billion commitment. OpenAI later announced a $30 billion Nvidia investment as part of its February 2026 funding round, a separate and smaller transaction from the original plan. With IREN, Nvidia did not simply buy an equity stake. IREN granted Nvidia a five-year right to purchase up to 30 million shares at $70, subject to regulatory and other conditions. The original OpenAI proposal and Nvidia's conditional investment right in IREN linked supplier capital to projects using Nvidia hardware, making independently funded demand harder to isolate. We covered that pattern in our explainer on AI circular deals.
The third is politics. The Gulf chip approvals are capped, conditional on US security rules, and can be pulled.
A sovereign deal, a hyperscaler contract and a neocloud's backlog can describe the same infrastructure spending from three different angles. Nvidia reports fiscal Q2 2027 results on August 26. Watch whether it updates the 80% sovereign-revenue growth rate or its footprint of nearly 40 countries.
Sources
- Related coverage: What is an AI neocloud?
- Related coverage: AI circular deals, explained
- Nvidia, fourth-quarter fiscal 2026 results (sovereign AI revenue)
- OpenAI, Introducing Stargate UAE
- OpenAI and Nvidia, strategic partnership announcement (letter of intent)
- OpenAI, Scaling AI for everyone ($30B Nvidia investment)
- Nvidia, NVIDIA and IREN strategic partnership announcement
- Oracle, fiscal Q4 2026 results
- L&T, teaming with Nvidia announcement
- Elysee, France AI investment announcement
- European Commission, InvestAI initiative
- US Department of Commerce, statement on UAE and Saudi chip exports (November 2025)
- CoreWeave, Q2 2026 results; Nebius, Q2 2026 results
This is general market commentary and opinion, not investment advice. Markets can go down as well as up, and you can lose money. Always do your own research and consider speaking with a licensed financial professional before making any investment decision.
