Industry·
Nebius's $40 Billion Bet Tests Neocloud Model Amid Delivery Delays
Nebius outlines aggressive 2027 capex plan financed by customer prepayments, but Microsoft contract delays and regulatory hurdles at its New Jersey site highlight execution risks for neocloud operators.

Nebius, the neocloud operator spun out of Yandex, used its third-quarter earnings call to unveil a $40 billion capital expenditure target for 2027 — a figure that far exceeds analyst expectations. The plan leans heavily on customer prepayments and debt secured by long-term contracts to fund data center construction before revenue fully materializes.
Financing Model Under Pressure
That strategy assumes timely delivery of capacity. However, the company's largest deal, a Microsoft agreement worth up to $19.4 billion for AI compute at a Vineland, New Jersey facility operated by third-party DataOne, is running months behind schedule. Only slightly more than half of the contracted capacity is now slated for year-end 2026, with the remainder pushed to the second quarter of 2027.
The delay stems from DataOne's accelerated shutdown of temporary natural-gas generators as the site transitions to Bloom Energy fuel cells. New Jersey regulators fined the facility $1.1 million after discovering at least 45 unpermitted gas generators in operation, and a noise-complaint lawsuit from nearby residents adds further uncertainty.
Microsoft's contract includes termination rights and financial penalties for missed deadlines. While the hyperscaler has not publicly canceled, reports indicate it is tightening oversight of third-party builds — including those from Nebius, IREN, and Nscale — and has already cited partner delays as a factor in its own lower-than-expected capital spending.
Why it matters for GPU / AI infrastructure: The neocloud playbook — rapid, debt-fueled buildouts backed by hyperscaler commitments — is being stress-tested. For GPU cloud buyers and investors, the episode underscores that contracted capacity dates are not guarantees, and that regulatory, permitting, and supply-chain realities can shift delivery timelines materially.
- aigpu
- ai gpu
- ai gpu cloud
- aigpu dubai
- nebius
- neocloud
- microsoft
- data-center
By AiGpu Editorial · Editorial rewrite based on public reporting (TechCrunch AI)
← All articles