Industry executives are currently watching a predictable shakeout unfold among the up-and-coming AI cloud providers known as neoclouds, and the numbers are finally catching up with the ambition. Australian provider Firmus abruptly pulled its IPO amid falling stock prices, while shares of CoreWeave and its smaller rivals dropped about 20 percent. Bloomberg reports that CoreWeave lost a quarter of its market value, while a measure of its credit risk hit a record high.
Investors are now forcing a reality check across the board. Some are pushing to cut British provider Nscale's target market cap from about $30 billion down to a much more grounded range between $16 billion and $19 billion. Meanwhile, CoreWeave CEO Michael Intrator has blamed smaller rivals like Firmus, Lambda, and Nscale for the current turmoil, noting in an employee memo that they are trying to raise enormous amounts of capital without the execution, revenue, scale, or software solutions that CoreWeave actually delivers to its customers.
Nvidia's Financial Exposure to Independent Providers
For anyone tracking the financial plumbing of the AI boom, the reliance on non-Big Tech customers is striking. Customers outside the major technology conglomerates account for 45 percent of Nvidia's data center revenue. This customer segment brought in $40 billion in the July quarter alone, more than double the figure from a year earlier. Because Nvidia's biggest traditional clients are busy designing and deploying their own proprietary silicon, the chipmaker has found itself forced to actively back its smaller partners with heavy financial scaffolding, including credit support and site acquisition assistance.
Nvidia provides a complex mix of equity investments, direct chip allocations, loan guarantees, and customer leads simply to prop up the demand side of the industry. A team led by Jensen Huang's son-in-law, Nico Caprez, helps neoclouds scout sites with available power, while Nvidia has also altered its loan guarantees to offer shorter terms for a smaller cut after larger neoclouds balked at its initial revenue-share demands.
Infrastructure Bottlenecks and Overcapacity Risks
Neoclouds are slamming into severe infrastructure barriers, led by missing power connections and increasingly gun-shy investors. Microsoft recently pulled out of a data center deal in Loughton, England, where it had planned to lease 23,000 Nvidia chips, simply because utility providers cannot guarantee power at the site until the 2030s. That is a glaring vulnerability when more than 40 percent of Nscale's $103 billion in contracted revenue depends directly on Microsoft.
The only thing that's going to happen with all these data centers is that prices will fall.
Gartner now predicts that 80 percent of independent neoclouds will cease independent operations by 2030, while Bernstein estimates that only 35 percent of the data center capacity announced in North America will ever actually get built. Microsoft CEO Satya Nadella stated on a podcast in late February 2025 that he expected an oversupply of compute starting in 2027 and that the industry was far too focused on speculative supply instead of actual, measurable demand.
For years, the market built its growth projections on a seemingly endless appetite for compute, operating under the naive assumption that every announced facility would fill up the exact moment the utility company turned on the power. As the financing tightens and power constraints bite, the bill for that overconfidence is finally coming due for both the neoclouds and their primary supplier.