Nebius Stock: The Gap Between a Signed Intent and a Powered Server

Nebius signed a letter of intent for Volt's Rotterdam AI campus, but binding contracts, funding and usable capacity remain unproven.

The most telling detail in Nebius's latest expansion news isn't the size of the campus. It's the distance between a handshake and a humming rack. A letter of intent signed yesterday for Volt's AI campus in Rotterdam opens a growth avenue, but it does not lock in a single megawatt of operational capacity — and that distinction sits at the fragile center of the entire AI infrastructure trade.

Demand can be announced. Infrastructure has to be financed, permitted, connected and built. That is a less glamorous story than a new gigafactory, yet it is the one that matters for anyone holding the stock.

A Rotterdam option, not a Rotterdam contract

Nebius is the first potential tenant of the planned Rotterdam site. According to a report by FD, the AI cloud company intends to take a substantial share of the projected capacity. The contract status, however, remains the crux: this is a declaration of intent, not a binding lease.

Volt is planning six buildings totaling up to 600 MW, with a projected overall investment of roughly EUR 7 billion. That figure belongs to the campus project as a whole. It is neither a disclosed contract value for Nebius nor a committed capital outlay by the cloud company. Grid connection and financing are no more secured in Rotterdam than the project's development itself, and no reliable operational timeline has been named. A potential anchor tenant can help a site along. It does not replace the funding or the power supply.

Two projects, two very different stages

The contrast with South Carolina sharpens the picture. There, Nebius and AIB signed a binding 50 MW contract on September 30, with an initial term of twelve years. Even that more advanced site is not yet running. Permits, financing and construction still lie ahead. According to AIB, customer prepayments are meant to cover part of the initial costs, while project loans and preferred equity financing had not yet been arranged.

Rotterdam sits a full contractual step earlier. Treating the two as equivalent available capacity would be a mistake. The real valuation question for Nebius is how much confidence growth deserves before the infrastructure behind it actually delivers. Binding contracts, secured funding and usable capacity are the milestones that will define the intent's economic worth — and they deserve more attention than the next headline campus number.

The market's mood is doing the talking

Reuters reported today on a selloff among AI suppliers and mounting doubts about the investment boom in the face of rising interest rates. Nebius itself lost ground, with a daily move of -2.8%. No clean read-through to Rotterdam can be drawn from that, but it fits a market that is growing more critical of expansion promises.

The broader skepticism has a specific edge for Nebius. Yesterday's decline reads less as a verdict on growth itself than as a vote of no confidence in how that growth gets funded. Worries about AI valuations and the capital-hungry data center buildout overshadowed even a positive assessment from Rosenblatt. That is understandable: a large order backlog creates opportunity, but it does not prove durably profitable capacity.

The daily move of -4.6% shows how sensitive investors have become to these doubts. At the same time, the stock is still up 156% year to date, meaning the pullback hit a share price that already carries substantial growth hopes. Reading the drop purely as sector weakness understates the company-specific financing question. Reading it as evidence that Nebius is failing ignores the reported business trajectory. The actual task lies between those extremes: weighing growth, investment needs and earning power together.

Revenue momentum meets depreciation reality

The numbers show why the debate cannot stop at top-line speed. In the second quarter of 2026, Nebius posted revenue of USD 582.3 million against depreciation of USD 259 million. Depreciation is not the same as cash going out the door, but it matters for judging earning power: computing capacity has to generate sufficient returns over its useful life. That is the real test of capital-intensive expansion.

The AI cloud segment reached USD 574.9 million in Q2 2026, a 514% jump year over year. That dynamism explains why additional capacity matters to the company. It does not automatically answer the questions about construction costs, financing and availability at new sites. Revenue growth and project risk can coexist; looking at only one of them makes the valuation too simple.

The reported backlog of remaining performance obligations stands at USD 37.5 billion, and Nebius has reaffirmed its 2026 revenue outlook of USD 3–3.4 billion. That argues against the blanket claim that the expansion lacks any commercial foundation. It also doesn't justify booking contractual commitments as profit. The backlog is a case for the growth story — not an all-clear on financing and profitability.

Pricing power versus project execution

Supply conditions appear tight, and the pricing reflects it. Media reports indicate Nebius raised GPU capacity prices again on October 1, with H100 rental rates up 17%. Back in May, the company had already lifted prices for that capacity by an average of 29%.

Higher rates can strengthen the economics of additional capacity. They are no guarantee that planned facilities come online on schedule. The bottleneck that supports today's higher rents is not proof that every future campus will be profitable.

Management is targeting annualized recurring revenue of USD 7–9 billion and an adjusted EBITDA margin of 40%. Those are ambitions, not achieved results. A margin target alone doesn't say what funds remain after investments. For shareholders, the decisive question is whether Nebius can finance its buildout without additional debt or further share sales eating into the growth advantage. That is a risk, not a foregone outcome.

By year-end, contractually secured power capacity is supposed to exceed 4 GW. The precise meaning matters here too: secured capacity is not the same as infrastructure that is already running and profitably utilized.

Insider sales, with the fine print attached

Adding to the pressure were share sales by the COO and CTO in early October. According to media reports, these went through prearranged trading plans — a detail that belongs in any honest reading before the transactions are turned into a verdict on the business outlook.

The sales are therefore no solid proof that management doubts the growth story. They did, however, land in a phase when investors were already questioning valuation and capital needs. That Rosenblatt's positive assessment failed to halt yesterday's decline fits that skeptical mood.

Media reports that OpenAI's annualized revenue came in below earlier market expectations added to the caution. No concrete revenue shortfall for Nebius can be derived from that. As a warning signal for return expectations across AI infrastructure, though, the debate is relevant — and it raises the pressure to back expansion plans with economic results.

What has to be proven

The re-rating of risk is justified. A premature write-off of Nebius is not. Revenue and backlog provide substance, and the growth perspective is convincing. The harder part of the story remains the financing and the sustainably achievable returns. Rotterdam adds an option; binding contracts, secured funding and usable capacity still have to give that option its economic value. Those are the milestones worth watching — more than the next big campus figure.