New cloud model cracks under billions of dollars of debt and looming mergers of AI warehouses
Our reading of liquidity movements in hyperscale computing warehouses reveals an accelerating fracture in an operating model that major AI labs have relied on to secure their compute capacity. In Silicon Valley and San Francisco, names such as CoreWeave and Nebius are no longer obscure infrastructure players; they have become the focus of heated financial debate after accumulating debt in the billions of dollars to build facilities packed with expensive compute hardware. These firms, known as the new cloud companies, purchase equipment up front and lease it to leading firms such as OpenAI, Anthropic and Meta, relying on long-term lease contracts as collateral for borrowing and expansion.
Kory Johnson, an analyst at The Quiet Period for IPO research, explains that this model requires full payment for equipment at the time of purchase, prompting operators to borrow against the contracts signed with customers as well as the AI chips themselves. He notes that lenders often look at the creditworthiness of lessee companies such as Microsoft and Meta more than they assess the operator’s own financial condition.CoreWeave, the sector leader, has accumulated debt of $35.1 billion because lenders pledged their facilities to the creditworthiness of its large customers rather than to its own solvency.As financial-market anxiety intensified, the company’s share lost a quarter of its value within a month, and the shares of competitors such as Nebius, Iron and Ablade Digital fell by similar or greater percentages, while debt investors began demanding much higher returns to cover the risk.
Most of these facilities trace their origins to the cryptocurrency-mining sector; CoreWeave was founded in 2017 to focus on mining Ethereum, and companies such as Core Scientific, Iron and Ablade Digital followed the same path before redirecting their physical infrastructure toward AI with the launch of the ChatGPT model in late 2022. The most notable exception was Nebius, which split off and changed its name in 2024 after having been part of the Russian search engine Yandex. Today the sector’s fate is entirely tied to the spending levels of technology giants and model developers; Melissa Auto, head of research at Visible Alpha within S&P Global, explains that the tech-arms race has injected massive spending into the ecosystem, yet the recent spending forecasts announced by Microsoft and Google fell short of Wall Street estimates, which was enough to push new-cloud-company shares lower.
This pressure coincides with clear operational setbacks; Microsoft cancelled an agreement with the British firm InScale amid energy-supply crises that could last until 2030, and Australian company Vermouth froze its plans to launch an IPO. Gartner forecasts that 80 percent of new-cloud companies will have to cease independent operations by 2030, while Charlie Day, vice-president and chief analyst at Forrester, believes the model faces growing financial risks that make an acquisition of at least three of these firms by larger players imminent, despite U.S. Treasury Secretary Scott Petersen telling CNBC that concerns were overstated and describing the sector as a small component of the overall system.Elon Musk warned that rising borrowing costs will push more than two hundred cloud companies toward a wave of collapse and mass consolidation, and that only a few will survive.
We see in these financial fractures a direct impact on infrastructure choices for those managing computing and technology systems in the Gulf, Egypt and the Levant. Regional development and digital-project teams rely on leasing flexible compute capacity to train models and run workloads without purchasing processors outright, but the confusion of independent operators under debt pressure creates two possibilities: either leasing costs rise as processor-rental fees incorporate banking-interest expenses for customers, or there is a risk of scheduled capacity being cut off if the provider defaults or is forced to sell. This scenario compels technology leaders in the region to review contract terms, devise contingency plans that avoid reliance on a single external provider laden with obligations, and to boost investment in local data centres and sovereign capacity to ensure operational stability insulated from Western credit volatility.
Even though Mike Intrator, CoreWeave’s chief executive, asserts that the market is moving irrationally and that major global firms are lining up for his company’s infrastructure services, and Yorkville IV analysts confirm that demand for compute still exceeds supply, the real challenge looms as investors await the pending IPOs of OpenAI and Anthropic. If this momentum slows, the side effects, according to Bradley Tusk of Tusk Ventures, will first hit the indebted warehouses with billions of dollars, turning the AI battle from a race for chips into a struggle to survive a debt guillotine.