AI companies have spent billions racing to build bigger data centers, but financing that expansion is getting harder. With borrowing costs at multi-decade highs, investors are questioning how much debt the industry can take on before the returns start falling short.
The selloff spreads
Twenty of the world's most valuable AI and semiconductor companies have lost more than $1T in combined market value this week, as concerns over the cost of funding AI expansion spread across global markets.
The selling hasn't been limited to Wall Street. SK Hynix and Samsung Electronics were among the Asian chipmakers hit hard, while European technology stocks also came under pressure.
Hotter-than-expected inflation sent the Nasdaq and S&P 500 to their worst single-day performance since July on Wednesday, raising the possibility of another Federal Reserve rate hike this year.
Treasuries find their footing
One unusual development came from the bond market. Government bonds rallied despite the hotter inflation reading, suggesting investors were more concerned about the stock market selloff than the immediate impact of rising prices.
That's a change from much of 2026, when inflation kept pushing yields higher and government bonds struggled to provide protection during market declines. Investors appear to be turning to Treasuries again, even as borrowing costs remain elevated.
Corporate debt is another matter. Bonds issued by heavily indebted AI infrastructure companies have come under pressure, sending yields to record highs and making future borrowing even more expensive.
The financing squeeze
Those higher costs are starting to affect actual projects. AI infrastructure companies are delaying fundraising or facing tougher terms as investors demand more convincing evidence that their investments will pay off.
Three planned listings have already been disrupted. Blackstone-backed AirTrunk's real estate trust, Nvidia-backed Firmus, and SoftBank-backed SB Energy have all been pulled or postponed.
AI Infrastructure Listings Pulled or Paused This Week
| Company | Backer | Deal | Status |
|---|---|---|---|
| AirTrunk REIT | Blackstone | Real estate trust listing | Pulled or paused |
| Firmus | Nvidia | $5B Australia IPO at up to $30.6B valuation (vs $10.5B in August) | Scrapped for a private round |
| SB Energy | SoftBank | Listing | Pulled or paused |
Firmus abandoned its planned $5B Australian IPO, choosing to pursue private funding instead. Its proposed $30.6B valuation was nearly three times the $10.5B valuation from an August funding round.
Investors were concerned about the company's debt and limited experience developing AI data centers. Reports that a major partner had withdrawn from an A$73B project only made the fundraising environment more difficult.
Not every company faces the same risks
The pullback doesn't mean AI investment is coming to an end. Investors are becoming more selective about which companies can finance their expansion and which ones depend on continued access to cheap capital.
Taiwan Semiconductor Manufacturing offers one contrast. The chipmaker reported its sixth consecutive quarterly profit record, although management also expressed caution about 2027 because of broader economic uncertainty.
Debt-heavy infrastructure operators face a tougher situation. CoreWeave has been among the companies under pressure, alongside Oracle, as investors reassess the cost of financing large data center projects.
Chipmakers haven't escaped the selling either. Nvidia, Micron, and Advanced Micro Devices traded lower ahead of Thursday's open, along with AI cloud provider Nebius.
What investors should watch
The biggest difference between these companies is how they're paying for growth. Businesses relying heavily on borrowed money face greater pressure as rates rise, while those funding expansion through their own cash flow have more flexibility.
That doesn't make profitable chipmakers immune to a slowdown. Their earnings still depend on customers continuing to spend, and TSMC's cautious comments about 2027 show how difficult it is to predict demand further out.
For the moment, another Federal Reserve rate hike would make an already expensive expansion even harder to finance, particularly for companies still waiting to turn their AI investments into meaningful profits.
