Nebius Stock: Higher GPU Rates, a 12-Year Capacity Deal and the Inference Bet Behind the Headlines

Nebius raised rates on selected NVIDIA GPU instances, won NVIDIA Exemplar Cloud status, and signed a binding 50 MW data center agreement.

Nebius has spent the past two weeks assembling the pieces of a larger commercial story: a binding infrastructure contract, repriced GPU instances, a certification from NVIDIA, and an acquisition aimed at strengthening its inference business. For investors, the central question is no longer whether the company can expand its offering, but whether that expansion translates into durable revenue.

New tariffs, but only for selected instances

As of October 1, published on-demand rates for certain NVIDIA GPU instances moved higher, according to media reports. The price list cites $8.50 per GPU-hour for the B200 and $9.50 for the B300.

The word "selected" carries weight here. This is not a documented price increase across the entire portfolio, and published rates are not the same as realized revenue. The change concerns the terms of the offering, not a reported jump in sales. Investors looking for proof of higher profitability will not find it in a price sheet alone.

On Thursday, Nebius also reached NVIDIA Exemplar Cloud status for training workloads on NVIDIA HGX B300. According to the company, NVIDIA validated the production infrastructure against its reference architecture. That recognition speaks to technical standing, while the tariff update speaks to commercial terms — two distinct signals that happen to arrive in the same week.

A 50-megawatt commitment with its own financing logic

On September 30, AIB Data Centers announced a binding agreement with Nebius for 50 MW of critical IT capacity at its CLT1 site in the southeastern United States. Media reports put the project budget at $800 million, including roughly $100 million in equity.

AIB later said expected customer prepayments are intended to cover that equity requirement. The distinction matters: this describes AIB's financing plan. Those prepayments should not be read as funds already received or as a confirmed payment from Nebius, and the project budget is not a disclosed capital outlay by Nebius itself.

The contract runs for an initial twelve years, with two five-year extension options attached. That horizon is a core feature of the deal and should shape how investors assess it — it can put expansion on steadier footing, but it also demands a view that extends well past short-term price moves.

Product and capacity are meant to work together

Nebius is not betting on capacity alone. Roughly a week ago, the company acquired Inferize, an inference-optimization specialist whose technology and development team were folded into Nebius Token Factory. Financial terms were not disclosed.

An October 1 cloud product update fits the same pattern. It includes Cloud Interconnect, which provides private connections between Nebius AI Cloud and customer networks — a concrete function that complements the infrastructure build-out.

The bull case, then, is not simply that more capacity equals more value. It assumes infrastructure, product features, and actual business usage reinforce one another. If that holds, the AIB contract could become a workable foundation for further growth, with the Inferize acquisition and the connectivity tools as supporting pieces rather than proof of economic success already achieved.

The bear case does not require a broken contract

The counterargument deserves equal weight. According to media reports, concerns about an AI bubble, higher interest rates, and heavy infrastructure spending drove broad selling pressure across the AI and neocloud sector on Friday. Coverage attributed the move in Nebius to the industry backdrop, not to any new company-specific announcement.

That leaves a risk no single contract can remove. Even if Nebius makes operational progress, the market could take a more cautious view of capital-intensive growth models. Higher rates and doubts about the economic viability of AI build-outs could push positive company news into the background.

The bearish thesis needs no assumed breach or failed expansion. It is enough that investors assign a lower weight to the expected benefit of additional infrastructure. Should economic utilization fall short of expectations, the long contract term could look less like a growth advantage and more like an open question about whether the build-out is justified by sufficient business.

Execution becomes the test

So long as added capacity and product improvements enable convincing commercial use, the growth scenario remains plausible. If confidence that infrastructure spending will be matched by corresponding business fades, the long-term agreement alone is unlikely to support the valuation.

The next concrete checkpoint is therefore the execution of the binding AIB contract — and investors should read its progress alongside evidence that the expanded offering is actually being used. Behind the headlines sit several separate steps: a binding capacity agreement, changed offering prices, and a completed technology integration. The common thread is the expansion of the commercial offering. Technical recognition and additional capacity remain distinct from their economic payoff — higher published tariffs on their own do not yet demonstrate higher profitability.