Little-known companies powering the artificial intelligence boom are attracting growing scrutiny as falling share prices, mounting debt and concerns over Big Tech spending raise questions about the sector’s financial stability.
A group of relatively obscure technology companies has found itself at the centre of growing concerns about the future of the artificial intelligence industry.
Known as neocloud companies, firms such as CoreWeave and Nebius operate massive data centres filled with expensive computing equipment, which they rent to major AI developers including OpenAI, Anthropic and Meta.
Although they have become increasingly important to the AI ecosystem, recent financial setbacks have raised questions about whether their business models are sustainable and what their struggles could mean for the wider technology sector.
Analysts at Gartner predict that by 2030, 80% of neocloud companies will have “cease[d] independent operations”. Meanwhile, experts at Forrester expect at least three companies in the sector to be acquired soon as financial pressures increase.
On Thursday, US Treasury Secretary Scott Bessent sought to downplay concerns, describing neocloud businesses as “a small component of the overall ecosystem” during an interview with CNBC.
Mounting debt raises concerns
Neocloud companies typically finance their operations through long-term leasing agreements with major technology customers that need immediate access to computing power.
These contracts, alongside the expensive AI chips purchased by the companies, can then be used as collateral to secure additional borrowing.
Cory Johnson, an analyst at IPO research outfit The Quiet Period, explained that debt is a common feature of the business because companies must pay for expensive hardware upfront.
CoreWeave, one of the industry’s leading players, illustrates the scale of the financial commitments involved. Despite being considerably smaller than some of the major technology companies it serves, it has accumulated $35.1bn (£26.3bn) in debt.
Johnson believes lenders may sometimes place greater emphasis on the financial strength of CoreWeave’s customers, including Microsoft and Meta, than on the company’s own creditworthiness.
However, investors appear increasingly uneasy about the sector’s financial exposure.
CoreWeave shares have fallen by approximately 25% over the past month. Other neocloud companies, including Nebius, IREN and Applied Digital, have experienced similar declines or even steeper losses.
Debt investors are also demanding higher returns from several companies in the industry, suggesting growing concern about the risks involved.
CoreWeave chief executive Mike Intrator has described the market reaction as “irrational” and expressed confidence in the company’s business model. He has also pointed to strong interest from some of the world’s most influential companies seeking access to its infrastructure and AI services.
From cryptocurrency mining to artificial intelligence
Several neocloud companies began their operations in the cryptocurrency mining industry, where businesses use powerful computers to generate digital currencies such as Bitcoin.
Because cryptocurrency mining and AI development require some of the same physical infrastructure, these companies were well-positioned to benefit when demand for artificial intelligence computing power surged.
CoreWeave, for example, was established in 2017 with a focus on mining Ethereum. IREN, Core Scientific and Applied Digital also have backgrounds in cryptocurrency mining.
Their existing infrastructure became particularly valuable following the launch of OpenAI’s ChatGPT in 2022, which helped accelerate investment and demand across the AI industry.
Nebius has a different history. It was previously known as Yandex NV and formed part of Russia’s version of Google before separating from the business and adopting its current name in 2024.
Today, the fortunes of these companies are closely tied to major AI developers and hyperscalers — large-scale data centre operators that are also among their most important customers.
Melissa Otto, head of Visible Alpha Research at S&P Global, described the enormous flow of investment into the sector as an arms race.
“There is just an enormous amount of spending getting injected into the system,” she said.
While neocloud companies have benefited from this investment, their dependence on major customers also exposes them to financial risks if those customers change their spending plans.
Forrester vice-president and principal analyst Charlie Dai said the business model came with “greater financial risk” when market conditions shift.
That vulnerability became apparent this past week after Microsoft and Google projected lower AI spending than Wall Street analysts had expected. Shares in several neocloud companies subsequently dropped.
The sector is also closely watching OpenAI and Anthropic, both of which are expected to begin selling shares to the public in the coming months.
Bradley Tusk of Tusk Ventures warned that if the AI market turns out to be a bubble and eventually bursts, neocloud companies would suffer significant consequences.
Musk warns of a wave of consolidation
The future of neocloud companies will also depend on their ability to continue borrowing money. Financing becomes more expensive as interest rates rise, potentially placing additional pressure on businesses already carrying substantial debt.
In a recent social media post, Elon Musk suggested that more than 200 neocloud companies could face what he described as “guppy-level death and consolidation”, with smaller businesses particularly vulnerable.
“Some will make it through, but not many,” he added.
Despite the concerns, some analysts believe demand for computing power remains stronger than available supply.
Analysts at Yorkville Ives said neocloud companies continue to fill an important gap in the market that larger data centre operators are not currently able to address.
Nevertheless, recent developments have added to uncertainty.
Microsoft recently cancelled an agreement with British neocloud company Nscale amid power supply problems that could continue until 2030.
Meanwhile, Australia’s Firmus has shelved plans to sell shares to the public.
These setbacks come as data centres face increasing public opposition over their environmental and community impact, alongside broader questions about how much artificial intelligence can improve worker productivity.
For Johnson, the industry’s financial challenges should not be dismissed.
“There’s reason to worry,” he said.
As investment in artificial intelligence continues to grow, the pressure on the companies supplying its computing infrastructure is becoming increasingly difficult to ignore. Their ability to manage debt, secure financing and maintain demand from major technology customers could play an important role in determining how sustainable the AI boom ultimately proves to be.
