Shares of major AI infrastructure providers fell on Monday after Elon Musk warned that more than 200 neocloud companies could face a wave of failures and consolidation, intensifying investor concerns about the debt financing an industry racing to meet demand for artificial intelligence computing.
Nebius, CoreWeave and Iren each fell more than 4% in early premarket trading before recovering some of their losses to trade around 2% lower. The declines added to a sharp sell-off over the past month, as investors reassessed whether the enormous sums being committed to AI data centers and computing equipment will generate returns sufficient to justify their financing costs.
“As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers,” Musk, chief executive of SpaceX and Tesla, wrote on X on Sunday.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
“That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many,” he added.
The warning comes from a business leader whose own company operates data centers and sells computing capacity, placing SpaceX in direct competition with neocloud providers. Musk’s comments therefore reflect both the financing pressures facing the sector and the competitive environment in which companies are seeking to secure customers and capital.
The underlying concern is that demand for AI computing may continue to grow while the financial conditions required to supply it become increasingly restrictive. Companies that have borrowed heavily to acquire expensive Nvidia chips and build data centers must generate enough revenue to cover operating expenses, interest payments, and the cost of capital. If customer demand fails to translate into sufficiently profitable utilization, even facilities equipped with advanced chips could become financial liabilities.
Musk was responding to a post by Chamath Palihapitiya, founder of Social Capital, who argued that the data-center expansion could leave the industry with substantial amounts of underused computing capacity by 2027.
Palihapitiya warned that there could be “a lot of partially stranded compute and compute sitting in inventory in 2027 than is currently forecasted”. Such capacity could be purchased and paid for but earn less than its total cost of capital, or remain idle.
“2027 will be a turbulent year because it will clear out a bunch of wrong/bad bets. It will also be a consolidation year for a few,” he said.
The concern goes beyond whether AI applications will attract more users. It centers on the chances of the financial returns from those applications to arrive quickly enough to support the infrastructure being built today.
Neocloud companies specialize in providing computing capacity, often using large clusters of advanced accelerators to serve AI developers and enterprises. Their business models can require substantial upfront investment in chips, data centers, electricity, and networking equipment before customers begin generating revenue at scale.
Borrowing can accelerate expansion when demand is strong, and financing is available. But it also increases financial risk when interest rates, credit spreads or lenders’ willingness to extend capital move against the borrower.
CoreWeave has become a focal point for those concerns. The sector’s sell-off intensified earlier this month after the spread on the company’s five-year credit default swaps rose to levels generally associated with financial distress. Yields on its bonds also climbed, signaling that investors were demanding greater compensation for the risk of holding its debt.
Credit default swaps are contracts that provide protection against a borrower defaulting. When their spreads rise sharply, it can indicate deteriorating confidence in a company’s ability to meet its obligations, although elevated spreads do not by themselves mean a default is imminent.
Over the past month, CoreWeave and Nebius have fallen more than 30%, while Iren has declined about 40%, according to the reported market figures.
The sell-off points to investors becoming more selective about which companies they believe can finance their expansion and convert computing capacity into sustainable earnings. It does not necessarily mean that demand for AI computing is weakening across the industry. Rather, the market is increasingly distinguishing between demand for the technology and the financial viability of the companies supplying it.
Nvidia and Nvidia, which has invested in several neocloud providers, sought to reassure the market at its GTC event in Berlin last week.
Neocloud Executives Defend Demand
Chief Executive Jensen Huang told CNBC that demand for neocloud services was “off the charts”. He explained that the main constraint was not customer appetite but access to the working capital required to expand capacity.
“What is missing in the middle is the working capital to fuel that growth… because there’s just so much demand across so many sectors of AI, there’s a scarcity of capital,” Huang said.
“The demand of the capital is much greater than the availability of the capital. That’s going to get solved,” he added.
Huang’s comments highlight the industry’s central contradiction. AI developers and businesses may be seeking more computing power, but infrastructure providers still need financing to purchase equipment and bring new facilities online. Strong demand does not automatically ensure that every provider can obtain capital on affordable terms.
Nvidia also has a commercial interest in the continued expansion of AI infrastructure. Neocloud providers are buyers of its chips, and their ability to finance new data centers affects demand for the company’s hardware. Its assessment of demand is relevant, but it does not independently establish that every neocloud investment will earn an adequate return.
Nebius, meanwhile, has sought to distinguish its financial position from CoreWeave’s. Marc Boroditsky, Nebius’s chief revenue officer, told CNBC: “We’re not carrying the kind of debt that they are.”
Boroditsky said Nebius expected to sell out the capacity it brings online in 2027 and suggested that a similar situation could emerge in 2028.
Iren Co-CEO Daniel Roberts also said customer demand remained strong despite the market’s growing scrutiny.
“We understand the increased focus from investors, but when we’re talking from customers, nothing has changed. The demand is still there,” Roberts told CNBC last week.
Roberts said discussions with customers for next year covered multiples of the capacity Iren could bring online, adding that conversations about 2028 had begun to gather momentum.
These statements indicate that the companies continue to see strong interest in future computing capacity. However, customer discussions and expected capacity sales are not the same as recognized revenue or cash flow. Investors must still assess contract terms, customer creditworthiness, equipment costs, construction schedules, and the margins available after financing expenses.
The Coming Test Is Profitability, Not Capacity Alone
The potential consolidation described by Musk and Palihapitiya would depend on how the industry balances its supply of computing capacity with customers’ willingness and ability to pay for it.
If AI adoption expands rapidly enough, providers with reliable financing, strong customer relationships, and access to power and equipment could benefit from sustained demand. Companies unable to raise capital on acceptable terms may struggle to complete projects, even where prospective customers are interested in their services.
The risk of overbuilding is equally important. Data centers and specialized AI hardware require significant upfront spending, and investments made today may take years to recover their costs. If several providers expand simultaneously on the assumption that demand will continue rising at current rates, the industry could eventually have more capacity than customers can profitably absorb.
That would create pressure on pricing and utilization, potentially weakening cash flow just as debt repayments and equipment-related expenses come due. Hardware also carries technological risks: the value of existing equipment depends partly on how quickly newer generations improve performance and efficiency.
The timing of supply and demand will therefore matter as much as their eventual size. Even if AI computing demand grows substantially over the longer term, companies can still face financial distress if their debt obligations mature before projects produce sufficient returns.
For now, the contrasting positions of Musk, Palihapitiya, Nvidia and the neocloud executives reveal a market divided over whether the current infrastructure boom is constrained primarily by financing or vulnerable to excess capacity.
The executives argue that customers are ready to absorb more computing power, while the bearish assessment warns that capital scarcity and potentially underperforming assets could force weaker companies out of the market.
Both outcomes can coexist. Demand for AI infrastructure can remain strong while individual providers fail because of excessive leverage, poor contract economics, or mistimed expansion.