The AI infrastructure boom has entered a new phase. Having access to GPUs is no longer enough. Companies must also prove they can finance expensive infrastructure, secure reliable customers, and generate enough cash to justify the investment.
SpaceX’s reported $40 billion debt financing deal highlights this divide. Elon Musk says the funding will support GPU purchases, while warning that more than 200 AI cloud providers could face consolidation or failure. SpaceX has an advantage that many smaller competitors lack: revenue from Starlink and rocket launches, alongside its AI business. That gives lenders a broader source of repayment than AI computing alone.
But this raises a bigger question: if the AI opportunity is so attractive, why are financing conditions becoming such an important risk?
1. CoreWeave: A huge backlog is not the same as financial security
CoreWeave ($CRWV) has become one of the most closely watched companies in the AI infrastructure trade. It provides access to powerful Nvidia GPUs for customers that need computing capacity without building everything themselves.
The opportunity is clear. AI developers need computing power, and CoreWeave has secured substantial customer commitments. Yet its business requires enormous upfront spending on chips and data centres, much of it supported by borrowing.
The problem is timing. Infrastructure costs arrive before all the associated revenue is collected, while interest must still be paid. If financing becomes more expensive, projects are delayed, or customers change their plans, the financial pressure can increase quickly.
CoreWeave’s reported backlog is important, but investors also need to examine contract duration, customer concentration, cancellation terms, and how much of that contracted revenue ultimately becomes free cash flow.
A large backlog can support future growth. It cannot automatically guarantee attractive returns.
2. Nebius: Funding access matters, but so does the cost
Nebius ($NBIS) has also been expanding its AI infrastructure footprint. Its reported $3.5 billion financing facility from Citi highlights the substantial capital required to build and equip data centres.
The key question is what this financing enables the company to earn relative to what it costs.
Investors should look beyond the headline funding amount. Is the financing secured against specific assets? What are the repayment terms? How quickly will new capacity become operational and generate revenue? Are customer contracts sufficiently strong to support the investment?
A large facility can help a company grow, but it also creates obligations. Expansion only creates lasting value when the returns from the infrastructure exceed its full cost of ownership and financing.
3. The balance-sheet test investors should watch
I would focus on five measures when comparing AI cloud providers:
• Free cash flow: Is the company moving towards generating cash after capital expenditure, or does expansion require another round of borrowing?
• Interest coverage: Can operating earnings comfortably cover interest expenses, even if growth slows?
• Debt maturity: Does the company face large repayments before its new data centres have had time to generate returns?
• Contract quality: Are customers creditworthy, contracts enforceable, and revenues diversified across multiple customers?
• Asset economics: How quickly do GPUs generate revenue, and will their earnings potential decline as newer chips arrive?
These measures help distinguish a company building a durable business from one that depends on continuously accessing fresh capital.
4. What could this mean for Nvidia?
The consequences extend beyond the cloud providers themselves.
Nvidia ($NVDA) benefits when customers invest heavily in AI computing. However, if smaller infrastructure companies lose access to financing, they may delay GPU purchases or reduce expansion plans.
That does not automatically mean AI demand will collapse. Spending could instead become concentrated among companies with stronger balance sheets, reliable customers, and better access to capital.
In that scenario, leading chip suppliers could still benefit from long-term AI growth, but the pace and distribution of infrastructure spending may change.
My takeaway
I don’t think the key question is whether AI computing demand will continue growing. It is whether every company building capacity can earn an adequate return on the money it spends.
SpaceX’s financing deal demonstrates that enormous sums can still be raised, but access to capital is not necessarily equal across the industry. CoreWeave and Nebius have different financial structures and growth plans, so they should be assessed individually rather than treated as identical bets.
For investors, a falling share price alone does not establish value. The more useful signal is whether the business is becoming financially stronger while its valuation falls.
In the next phase of the AI boom, the companies that survive may not simply be those with the most computing power, but those that can turn that capacity into sustainable cash flow.
For discussion purposes only. This is not financial advice.
Comments