AI Debt Bubble: CoreWeave Bonds Hit 80 Cents, IPO Market Breaks

The AI trade has a dirty little secret, and it lives in the bond market. While Jensen Huang is out here comparing today's AI boom to the early internet and everyone's busy debating Nvidia's next chip cycle, bond investors are quietly pricing in something much uglier. CoreWeave ($CRWV) bonds are now trading below 80 cents on the dollar with yields above 15%, which is not a sign of a healthy company. That is a distress signal wearing a neon sign.

The Bloomberg Daybreak team laid out the math clearly: US semiconductor stocks erased $1.6 trillion in market cap over six sessions, CoreWeave shares fell 17% in a single day (its worst drop in over seven months), and the neocloud model of funding data center buildouts entirely with high-cost debt is getting its first real stress test. BondCliQ's Chris White flagged that AI-linked companies have issued approximately $258 billion in bonds, many of which are now trading significantly wider just six months after issuance. That's not noise. That's a structural crack.

Meanwhile, across the Pacific, Firmus, an Australian data center operator backed by Nvidia, watched its highly anticipated ASX IPO collapse within 48 hours of its book build closing in July 2026. The deal was seeking to raise over $5 billion at a $30 billion valuation, which was a jaw-dropping jump from its last private round at $10.5 billion. Institutional investors took one look at that math and walked. Then ran. The Firmus implosion is a useful data point because it shows that public market investors, unlike the venture crowd, are actually doing the math on AI infrastructure valuations right now.

Warburg Pincus CEO Jeffrey Perlman added more context this week, noting that the IPO market is structurally broken with 60% fewer public companies than a decade ago, and that AI-related IPOs are being pulled or delayed as parts of the AI complex trade at distressed levels. His bigger concern: 10 companies represent 50% of the S&P 500, and that concentration is pushing serious institutional money toward private assets. The public market is becoming a one-big-bet casino, and some players are cashing out their chips.

To be fair, Goldman Sachs APAC strategist Timothy Mo thinks the market is overcorrecting on AI supply chain stocks, pointing to strong ASML order data and token usage doubling in nine months as reasons the demand story is still intact. And Meredith Whitney's counter-argument, that a pullback in AI capital spending could trigger a real economic slowdown given how much GDP growth is being propped up by that capex, keeps the bear case honest too.

The bull and bear cases are both defensible right now, which is exactly what makes this moment so uncomfortable: the AI trade is no longer a consensus, and the bond market usually figures these things out before anyone else does.