These firms you've never heard of are feeding fears the AI bubble could burst

Neocloud firms often finance themselves by signing long-term lease contracts with major tech customers who need quick access to computing power.Firms can then borrow money using those contracts – plus the AI chips they purchase – as collateral, explained analyst Cory Johnson of "The Quiet Period" IPO research outfit.He said debt is common across neocloud firms because the hardware has to be paid for up front.Take the industry leader: CoreWeave. It's a fraction of the size of some of the Big Tech companies it serves.But Johnson believes lenders sometimes consider the creditworthiness of CoreWeave's customers, like Microsoft or Meta, more than the company's when they lend."This helps explain how CoreWeave accumulated $35.1bn (£26.3bn) in debt," he said.Some jittery investors appear to be having second thoughts.CoreWeave stock has lost a quarter of its value over the past month, and it's not alone.Shares in other neocloud firms like Nebius, IREN, and Applied Digital have all shed about that much if not more during the same period.Debt investors also appear spooked, and are demanding ever-higher returns from several firms in the sector.CoreWeave's CEO Mike Intrator has said that the market is acting "irrationally" and has expressed confidence in its business model. Some of the world's most consequential companies are queuing up for CoreWeave's infrastructure and other AI services.Several neocloud firms have their roots in the volatile crypto mining industry, made of companies that generate cryptocurrencies, like Bitcoin.Crypto miners use some of the same kind of physical infrastructure that is used to develop AI.CoreWeave was founded in 2017 with a focus on mining Ethereum.IREN, Core Scientific and Applied Digital also started in crypto mining.Their infrastructure left them well-positioned when AI exploded in 2022 with the introduction of OpenAI's ChatGPT.One outlier is Nebius, which was known as Yandex NV when it was part of Russia's version of Google before splitting off and changing its name in 2024.Now, the success of these firms hinges on the fate of major AI developers and hyperscalers – even larger data centre operators which neocloud firms often count as key customers."There is just an enormous amount of spending getting injected into the system," said Melissa Otto, head of Visible Alpha Research at S&P Global. "It's an arms race."And neoclouds have benefitted.But the trade-off of their model is "greater financial risk" in the face of change, said Forrester VP and principal analyst Charlie Dai.That includes any adjustments that their Big Tech customers might make to their spending plans.This past week, when Microsoft and Google forecast less AI spending than Wall Street analysts had projected, shares in several neocloud firms tumbled.Much is also riding on Anthropic and OpenAI, which are both expected to start selling shares to the public in the months ahead.If AI proves to be a bubble – and bursts – neocloud firms "are definitely going to face collateral damage," said Bradley Tusk of Tusk Ventures.The future of the neoclouds also depend on their ability to keep borrowing, which gets more expensive as interest rates rise.In a recent social media post, Elon Musk suggested this challenge could lead more than 200 neoclouds to "see guppy-level death and consolidation", with the smallest firms being the most vulnerable. "Some will make it through, but not many," he added.For now, analysts at Yorkville Ives said demand for computing power still exceeds supply, adding that neoclouds are filling an important gap that larger data centre companies cannot fill.But to some, the neocloud sector's recent wobbles are concerning.Microsoft recently cancelled a deal with a British neocloud outfit called Nscale amid power supply issues which could last through to 2030.Australia's Firmus just shelved plans to start selling shares to the public.These developments come amid a public backlash against data centres and questions about how much value AI can add to worker productivity."There's reason to worry," said Johnson.