Why Nebius Needs $250 Billion to Scale Its AI Cloud Through 2030

Europe's leading neocloud has sold out its GPU capacity through 2027 and posted Q2 2026 revenue up 454% year-over-year, but its disclosed $250 billion capital requirement through 2030 reveals a funding ambition that dwarfs everything the company has raised to date.

Nebius Group indicated this week that executing its AI data center and compute expansion plans through 2030 will require $250 billion in capital — a figure that crystallizes the central tension of the neocloud moment: demand for AI infrastructure is real and contracted, but funding it at the scale now being contemplated is testing the limits of what capital markets can absorb. The disclosure arrives just weeks after the Dutch-registered company delivered one of the most striking operating quarters in the sector's brief history, turning acute market skepticism about a convertible bond offering into a confirmation that Nebius is not merely riding the AI wave but actively shaping its supply side.

The Q2 2026 results, released in August, showed AI cloud revenue of $582 million, up 454% year-over-year, with adjusted EBITDA turning positive at $236 million — a reversal from a $21 million loss in the year-earlier period. Single-quarter capital expenditure reached $5.66 billion, already exceeding the company's entire fiscal year 2025 capex of $4.07 billion (itself up 403% from $808 million in fiscal 2024, per Slicast's compilation of SEC XBRL filings). Against FY2025 revenue of $530 million, that translates to a capex intensity of 767% — the highest among the seven publicly tracked neoclouds in Slicast's coverage universe. On the commercial side, a Microsoft Azure supply agreement for up to $19.4 billion in dedicated GPU capacity, combined with a Meta anchor contract and four additional Q2 signings each exceeding $1 billion, pushed total contract backlog past $46 billion. Capacity is fully booked through 2027.

The company's capital formation strategy has evolved in step with its commercial wins. In late August, Nebius priced $5 billion in convertible senior notes — upsized from an initial $4.5 billion target — a raise that initially sent shares sharply lower before the market reassessed the deployment rationale and partially recovered. That bond issuance, layered onto the Microsoft agreement and a new strategic partnership with Palantir to serve as its sovereign AI infrastructure provider, traces the arc of a company that entered 2025 as a post-Yandex restructuring story and has since built a contract base large enough to anchor multi-billion-dollar debt instruments. Jensen Huang said in early September that a one-gigawatt AI data center facility is worth $50 to $60 billion — a valuation framework that, if it holds, materially backstops the economics of Nebius's expansion sites in New Jersey, Wales, and communities across the United States where zoning approvals and tax abatement negotiations are already underway.

The risks are multilayered. A $250 billion capital requirement through 2030 implies a sustained annual funding rate that would far exceed the company's recently completed $5 billion convertible raise; whether bond markets, equity markets, and customer prepayments can collectively absorb that quantum over four years remains genuinely uncertain. Meta's announced plans to scale its own compute infrastructure introduced a meaningful competitive risk for neoclouds whose revenue depends partly on hyperscaler outsourcing — shares of Nebius, CoreWeave, and IREN fell in tandem on that news in early September. Interest rate sensitivity has been demonstrated in practice: a rate-pressure episode in August triggered a roughly 5% single-session decline in Nebius shares. On the ground, voters in Independence, Missouri recalled a city council member in September after he approved $6 billion in tax incentives for a Nebius data center facility — a signal that community and political resistance to AI infrastructure siting is evolving into a material permitting risk. And while Q2 adjusted EBITDA of $236 million confirms the business model works at current scale, a 767% capex-to-revenue ratio is structurally unsustainable; the path to balance-sheet self-sufficiency requires either revenue to compound dramatically faster than capital expenditure, or external capital to bridge the gap across multiple additional funding cycles.

Three signals merit close attention in the coming quarters. First, whether Nebius can extend its anchor customer base beyond Microsoft and Meta to additional hyperscalers or sovereign buyers — a more diversified demand base would underpin larger and less concentrated capital raises. Second, the trajectory of quarterly EBITDA margin: the Q2 result is a genuine inflection, but widening that margin while single-quarter capex runs above $5 billion is the defining operational test, and investors will watch for evidence that early deployment of NVIDIA LPX inference racks and fully contracted 2027 capacity translate into durable pricing power. Third, the evolution of customer prepayment structures — if anchor customers can be persuaded to pre-fund capacity at scale, as has been reported to be under discussion, that shifts a meaningful share of the funding burden off Nebius's balance sheet, reduces equity dilution risk from future raises, and tightens the economics of the $250 billion plan considerably. The company has compressed into roughly twelve months the commercial credibility that most infrastructure businesses take a decade to establish. Whether the capital markets will remain patient enough to let that credibility compound into cash-flow self-sufficiency is the question 2027 will begin to answer.