Liz’s view
The bottleneck in the AI boom was chips, then power, then memory. It’s about to become money.
That might sound strange as Anthropic prepares for a $1 trillion-plus IPO and data centers keep sprouting like big-box poppies. But the fear and exasperation among the lenders whose money fuels this game sharpened noticeably over the past week.
It’s not just the days-long chip selloff that’s messing with global markets. “There’s shortages in chips. There’s shortages in power. And soon, there will be shortages of capital,” the CEO of Singapore’s biggest bank, DBS, warned Semafor’s Shelly Banjo in Bangkok this week.
“We’re not an ATM for your dystopian future,” a major financier of data centers told me yesterday. Rest assured, he would happily play the part if the risk-return math made sense, but it doesn’t. Good luck enforcing Potemkin guarantees from hyperscalers or arguing that local protests excuse a construction delay that blew a hole in the financing.
Investors get scared the same way moguls go bankrupt, gradually, then quickly. You see it in deals like Volta’s $5 billion Norwegian data center, which months ago would have been waved through by investors now demanding tougher terms. You see it in CoreWeave’s tanking bonds. You see it in the cost of a policy to protect against SpaceX defaulting on its debt, which goes for 80% more than it did three months ago. And in the early price talk for the rocket company’s $40 billion loan being marketed this week, which I’m told is around 1 percentage point higher than its existing bonds. To quote Elon Musk himself: Yikes.
Yes, OpenAI will have no trouble raising the $30 billion it’s looking for and equity markets will probably recover from this week’s chip selloff. But it’s worth remembering that telecom bonds began cracking six months before the dot-com crash. In 2007, subprime mortgage bonds peaked eight months before the S&P 500 did. Equity investors get paid to be dreamers. Lenders get paid not to lose money.
Notable
- For now, spending on the AI remains strong: Goldman Sachs expects AI infrastructure investment to drive a $7.6 trillion boom through 2031.