Three AI listings stalled in a month: the money behind the AI boom is getting pickier

The squeeze is likely to widen the gap within the AI trade. Cash-generating hyperscalers and established suppliers can fund spending internally, while newer, debt- or equity-reliant players face higher costs and a narrower window. Credit is the market to watch: spreads on high-yield AI infrastructure debt, led by CoreWeave, are the earliest warning of whether stress is spreading or staying contained. In Asia, Korean and Taiwanese chip stocks are highly sensitive to any slowdown in AI capital spending, a risk already reflected in SK hynix’s roughly halved market value since June. Oil is part of the cost-of-capital story: Brent above $110 a barrel keeps inflation and rate expectations high, keeping bond yields elevated and raising the hurdle for capital-intensive projects whose payback is years away.--- Demand for AI hardware remains strong, but investors are becoming much more selective about how they fund it, and the companies most dependent on outside capital are feeling it first.Summary:Three AI-linked listings have been cancelled, halted or delayed this month: Firmus Grid’s roughly $5 billion IPO, Airtrunk’s Singapore listing and SK hynix’s planned Solidigm IPO.OpenAI is reportedly near a funding round of up to $30 billion at a $1.25 trillion pre-money valuation, below the roughly $1.4 trillion it initially sought.Debt markets are flashing stress, with yields on some of CoreWeave’s unsecured debt around 15%.The Philadelphia semiconductor index fell 6.7% on Thursday, its worst day since June, as Meta’s capital spending renewed doubts about returns on AI investment.Counter-evidence remains strong: JPMorgan says positioning is largely washed out and the tech sector is generating about $400 billion more in cash than it spends, while Western Digital reported surging demand and prices.Upcoming tests include megacap earnings, the Fed decision and SK hynix results on October 28, OpenAI’s funding round and Anthropic’s expected listing in mid-November. The AI boom has not run out of demand, but it is running into a more cautious funding market. In the space of a few weeks, three planned listings linked to AI infrastructure have been cancelled, halted or delayed, private valuations have been marked below earlier targets, and parts of the debt market are pricing in genuine distress. Together, they suggest that investors who financed the build-out with few questions are becoming more selective.The latest sign came on Friday, when SK hynix said its Solidigm unit had ended a review of a US initial public offering intended to help fund a new fabrication plant in the US. The South Korean chipmaker cited market conditions and share price trends, and said the decision did not reflect any change in Solidigm’s growth prospects. Reports earlier this autumn had suggested the listing could value the unit at up to $150 billion.Solidigm followed two similar decisions in Asia-Pacific. Earlier this month, Australian data centre operator Firmus Grid cancelled an IPO of almost $5 billion that would have been the country’s largest in nearly 30 years, citing volatility. Days later, Blackstone-backed Airtrunk pushed a planned Singapore listing into the first half of next year, The Wall Street Journal reported.The repricing extends to private markets. OpenAI is close to finalising a funding round of up to $30 billion at a pre-money valuation of $1.25 trillion, led by investors from the United Arab Emirates, according to reports this week. Earlier in October, Bloomberg reported the company had presented a valuation of about $1.4 trillion to investors as a fixed price. The new figure is still far above the roughly $850 billion valuation from March, but it suggests investors have pushed back on price. Reuters reported that OpenAI is rushing to secure commitments before rival Anthropic’s expected stock market debut in mid-November, as both draw on a similar pool of capital.The strain is most visible in credit. CoreWeave, the largest of the neoclouds that rent computing power for AI workloads, has seen yields on some of its unsecured debt climb to around 15%, a level usually associated with distress, after heavy borrowing to build data centres. Analysts are divided. HSBC has flagged structural concerns across the neocloud sector, while Goldman Sachs argues CoreWeave’s use of asset-backed project financing limits the rise in its borrowing costs.Public markets have reflected the same doubts. The Philadelphia semiconductor index fell 6.7% on Thursday, its biggest daily drop since June, and the Nasdaq slid more than 2.5% after earnings from major technology companies, including Meta’s spending plans, renewed concerns that the returns from AI may not justify the investment. SK hynix’s market value has roughly halved since its June peak.Several forces are behind the change. Interest rates are rising: the Federal Reserve resumed hikes in September and the Bank of Japan has pushed its policy rate to a 31-year high. Higher rates make it more expensive to fund projects whose revenue arrives years after the spending. Geopolitical risk has also increased, with the war involving Iran keeping oil above $110 a barrel and inflation expectations elevated. And as the scale of AI spending has grown, investors have become more focused on when it will pay off.There is a strong counter-argument. JPMorgan strategists said this week that hedge funds have already cut much of their exposure to chip stocks, which could limit further downside. They also noted that neocloud capital spending of about $80 billion this year is small next to roughly $1.8 trillion of total technology investment, and that the broader tech sector is generating about $400 billion more in cash than it spends, giving larger companies room to absorb any funding gaps. Demand also looks real: Western Digital said this week that revenue rose about 55% as AI customers drove a 28% rise in storage prices, and that it has demand visibility to 2030 or 2031 from its largest customers. OpenAI has told investors its annualised revenue has reached about $50 billion.Taken together, the evidence points to a split rather than a collapse. Companies that can fund AI spending from their own cash flow remain well placed. Those that depend on external capital, especially debt-funded newcomers and new listings, face higher costs and tougher terms. That is a healthier sign of discipline than a reckless boom, but it also means growth may slow at the margin for the most leveraged parts of the supply chain.Several tests are coming. Megacap earnings will show whether spending plans hold. The Fed decides on October 28, the same day SK hynix reports results. OpenAI’s round has yet to close, and Anthropic’s listing next month will show whether public investors are still willing to back AI at scale. A strong reception would suggest the window is reopening. Further pulled deals or wider spreads on AI-related debt would indicate the squeeze is deepening.   This article was written by Eamonn Sheridan at investinglive.com.