The warning came from the bond market

The most revealing development in financial markets this week came from an unexpected corner of the artificial intelligence boom. Investors have spent months debating whether AI companies can justify their enormous valuations. Now the bond market is asking a more immediate question. Can they afford the infrastructure being built to support those valuations?

CoreWeave became the focus of that concern. Shares in the AI cloud provider plunged more than 15% on Wednesday as its borrowing costs surged. The yield on one of its bonds briefly reached 16%, while the cost of insuring its debt against default exceeded 1,100 basis points. These are levels associated with considerable financial stress.

CoreWeave has accumulated approximately $50 billion in debt, according to Reuters, to finance its rapid expansion. Management argues that its substantial contracted revenue backlog provides sufficient coverage for existing obligations. Nvidia has also committed to purchasing unsold computing capacity under an agreement extending into 2032.

Those arrangements provide meaningful support, but contracted revenue is not the same as free cash flow. Data centres require expensive processors, electricity, cooling and financing. Technology also becomes obsolete, creating recurring investment requirements.

The concern is therefore less about whether AI demand exists than whether the companies supplying that demand can earn adequate returns after financing their expansion.

Inflation makes the financing problem harder

The timing of this reassessment was particularly uncomfortable.

Wednesday's US consumer price index showed inflation accelerating to 3.7% in September from 3.4% in August. Monthly prices increased 0.6%, with gasoline accounting for more than half the advance. Core inflation, excluding food and energy, reached 2.6% annually.

Thursday brought another warning. Producer prices jumped 1.3% in September, their largest monthly increase since March 2022, taking annual producer inflation to 6.3%. Although energy contributed heavily, rising services costs suggested broader pressure on business expenses.

Geopolitics remains central to the problem. Brent crude settled at $105.92 a barrel on Thursday as investors weighed potential sanctions relief against continued disruption around the Strait of Hormuz. Sustained energy costs threaten inflation expectations and corporate margins, complicating the Federal Reserve's ability to ease financial conditions.

Markets responded sharply. The Nasdaq fell 1.88% on Wednesday, while semiconductor stocks extended their losing streak. Thursday's selling spread across Asia, where South Korea's KOSPI dropped almost 8% and Japan's Nikkei nearly 5%.

Interestingly, US Treasury yields fell despite the inflation surprise as investors sought safety. The benchmark 10-year yield declined approximately 13 basis points on Thursday to 5.064%. Wall Street also recovered from early losses, with the S&P 500 gaining 0.33% as investors moved towards healthcare and other defensive sectors.

That divergence is important. Investors were no longer simply selling equities because interest rates might rise. They were beginning to distinguish between companies capable of financing growth internally and those dependent on increasingly expensive external capital.

Strong demand does not guarantee attractive returns

The contrast with Taiwan Semiconductor Manufacturing Company was striking.

On Thursday, TSMC reported third-quarter revenue growth of 50.9% in Taiwan-dollar terms and a 68.7% increase in net profit attributable to shareholders. Its gross margin reached 66.4%, while management projected fourth-quarter revenue of $52.3 billion to $54.3 billion.

These figures demonstrate that substantial profits are already being generated within the AI supply chain. The difficulty is that profitability is unevenly distributed.

Semiconductor manufacturers can earn attractive margins selling the equipment required for AI development. Infrastructure operators must then recover the cost of that equipment through years of computing services, while managing depreciation, financing expenses and technological change.

Over the next 6–24 months, investors should pay particular attention to the relationship between capital expenditure, operating cash flow and debt servicing. Credit spreads, refinancing terms and the conversion of contracted revenue into cash will become increasingly important indicators of financial resilience.

There is also a broader market implication. If lenders demand higher returns from AI infrastructure projects, some investment plans may become uneconomic. That could eventually affect semiconductor orders, data-centre construction and the earnings expectations supporting technology valuations.

The alternative remains plausible. Strong computing demand, improving utilisation and productivity gains could generate sufficient cash flow to justify today's investment commitments. TSMC's results provide substantial evidence that the underlying demand is real.

But this week's market reaction exposed a distinction that investors can no longer comfortably overlook. A transformative technology does not automatically produce a profitable investment at every point in its supply chain.

The AI revolution may continue delivering extraordinary technological progress. Whether every company financing that revolution can deliver equally impressive financial returns is becoming a much more demanding question.

Sources and references

• reuters.com
• reuters.com
• bls.gov
• bls.gov
• investor.tsmc.com
• reuters.com