Neocloud Shakeout Looms, and Nvidia Has a Lot to Lose
Phoebe Liu reports that after a week in which Firmus abruptly pulled its IPO and shares of CoreWeave and smaller rivals fell around 20%, data-centre and neocloud executives expect consolidation: two executives said only seven to 10 neoclouds will remain in a few years, versus the hundreds now trying to get going. Cirrascale CEO David Driggers expects asset managers and telecom/fiber companies to pick up neoclouds and says Cirrascale is free-cash-flow positive with more than $200M in annual revenue; another executive thinks consolidation could start by late next year, with CoreWeave and Nebius as possible acquirers. Gartner predicts 80% of independent neoclouds will cease independent operations by 2030; Bernstein estimates only 35% of announced North American data-centre capacity will be built (via Bloomberg). Nvidia's 'everyone else' customer bucket, which includes neoclouds, was $40B, or 45% of data-centre revenue, in the July quarter, more than double a year earlier. Nvidia has signed $20B of its own data-centre leases to pass on to customers, has a team led by Nico Caprez and Raj Mirpuri helping neoclouds find powered sites, and now offers shorter-term credit support for a smaller revenue cut after some bigger neoclouds declined its initial offer (per one executive). The key distinction: a lessee like CoreWeave could fail without derailing the buildout, but builders like Nebius, IREN and Nscale would hurt Nvidia more. Caveats: consolidation forecasts are executives' and analysts' views; the Anthropic over-commitment claim is one data-centre power investor's belief; the piece says lenders would probably renegotiate rather than let a cloud go bankrupt.
Why it matters
Neoclouds have absorbed huge amounts of private capital, and Europe has several of the 'builders' named as most exposed (Nebius in Amsterdam, Nscale in the U.K.). A forecast of hundreds shrinking to 7–10 is a direct prompt to track debt, Nvidia backstops, IPO pulls and M&A across the category, and to watch for distressed sales and roll-ups.
Executive takeaways
- Consolidation call: two executives see only 7–10 neoclouds surviving in a few years out of hundreds; one sees it starting by late next year, with CoreWeave and Nebius as possible buyers and asset managers or telecom/fiber firms as others.
- Lessee vs builder: CoreWeave mostly leases, so its sites could find new tenants; Nebius, IREN and Nscale build their own, so their capital trouble could stall the buildout and Nvidia's sales.
- Nvidia exposure: the non-hyperscaler bucket was $40B (45% of data-centre revenue) in the July quarter, more than double a year earlier; Nvidia holds $20B of its own leases and has a $6.3B rent-back backstop for CoreWeave (as of Sept 2025).
- Credit support is shifting: Nvidia now offers shorter-term support for a smaller cut of revenues after some bigger neoclouds declined the first terms (one executive); Huang said CoreWeave won't need a further backstop.
- Who carries the risk: guarantors such as Nvidia, Broadcom, Google or AMD could be left on the hook and might run the facility or find a tenant.
- Third-party skepticism: Gartner sees 80% of independent neoclouds losing independence by 2030; Bernstein sees only 35% of announced North American capacity getting built.
- Demand-side risk: one investor thinks Anthropic deliberately over-committed to leases in case some sites fall through, and one neocloud executive called eventual oversupply a given.
What The Information may be missing
The piece doesn't size the category: no count or list of the 'hundreds' of neoclouds, no total debt figure, and no breakdown of how much Nvidia has committed across equity, leases and guarantees beyond the CoreWeave backstop and the $20B of leases. It is U.S.-centric, with little on European or Asian players or sovereign-AI demand that could keep regional neoclouds alive. It doesn't bring in the same-day news that Microsoft pulled out of Nscale's Loughton site, a concrete example of the builder risk it describes. Cirrascale's revenue and cash-flow claims come from its CEO and aren't independently checked, and the consolidation timelines are forecasts, not reported deals.
