The AI cloud is built on a mountain of debt

The financial underpinnings of the AI infrastructure boom are looking increasingly precarious. The latest financial results from CoreWeave and Nebius, two of the most prominent 'neocloud' companies renting out sought-after GPU capacity, paint a stark picture of a sector fuelled by enormous debt and customer advances rather than operational profit. CoreWeave, for instance, saw its revenue more than double year-on-year to nearly $2.6 billion, yet its operating expenses were higher, leading to an operating loss of $49 million and a net loss of $626 million for the quarter. This performance comes against a backdrop of staggering financial commitments, including total indebtedness of $35.6 billion and planned capital expenditures for 2026 between $35 billion and $39 billion.

Nebius is following a similar script of aggressive, debt-financed expansion. The company announced ambitious plans to bring over a gigawatt of new data centre capacity online every year from 2027, an endeavour that will require between $20 billion and $25 billion in capital expenditure this year alone. Like CoreWeave, Nebius is not yet profitable, posting a $176 million operating loss in the second quarter. To fund its build-out, the company is relying on asset-backed debt facilities, using its GPUs as collateral, and is on track to collect over $9 billion in customer prepayments this year. This effectively means customers are providing unsecured financing for their supplier's expansion, betting on its long-term survival to see a return on their down payments.

This high-leverage model introduces a significant systemic risk into the AI supply chain. These companies are the critical link between hardware manufacturers and the thousands of enterprises scrambling for AI compute. Their financial health is therefore not merely an internal matter but a key variable in the stability of the entire ecosystem. The risk is compounded by high customer concentration. At CoreWeave, just three customers accounted for 72 percent of its quarterly revenue. A change in strategy from even one of these major clients could have severe consequences for the provider's financial stability, with potential knock-on effects for its smaller customers.

For finance leaders, this dynamic fundamentally changes the nature of supplier due diligence for AI infrastructure. The sector's revenue growth, while impressive, masks underlying vulnerabilities. A CFO cannot simply take a provider's market position at face value; a deeper analysis of its balance sheet, cash flow, debt covenants, and customer dependencies is essential. Relying on a single, highly leveraged provider for mission-critical AI workloads is a high-stakes gamble. The potential for service disruption, sudden price hikes to service debt, or even supplier insolvency is very real. The strategy of using customer prepayments to fund capex should be a major red flag, requiring finance teams to assess whether they are comfortable acting as an unsecured lender to a key supplier in a volatile market.