Finance podcasts are doing the AI math: debt risk on one side, OpenAI's revenue story on the other

The credit scare, per Fast Money

On CNBC's Fast Money, Citi head of equity trading strategy Stu Kaiser argued the AI infrastructure trade is developing a credit problem, pointing to CoreWeave. The company's unsecured debt due in 2032 is trading at yields above 15 percent, Kaiser noted; CoreWeave allocates 25 percent of its revenue to debt service, faces a short maturity profile requiring refinancing within five to seven years, and accounts for 5 percent of the $258 billion in debt issued by AI-linked companies over the past 12 months. Kaiser characterized the AI-linked debt as potentially subprime-tranche in nature, the episode's summary said. m:…2186

The panel widened the lens beyond one borrower. Dan Nathan warned: "If the demand does not materialize, then these contracts are going to go away." Oracle's credit default swaps widened by 25 basis points to 288 — higher than levels seen in the summer of 2008, per the episode — and BlackRock's Larry Fink said that if there were a moratorium on data-center construction in the United States, "we will be in a recession." The through line of the segment: the AI build-out is no longer being priced purely on growth. m:…2186

Friar's rebuttal: the demand is already here

OpenAI CFO Sarah Friar, speaking to Bloomberg, pushed back on the bear case with revenue numbers. OpenAI hit close to $50 billion in annualized revenue in Q3, she said, and the company expects to reach at least $70 billion for the year. She added that more than half of the tokens OpenAI sees in Europe are now agentic — "we've gone over 50% of all of the tokens we see are agentic" — and pointed to usage elasticity: after cutting prices on the Luna model series by 80 percent about six weeks ago, "We saw use go up 10x." m:…7518

Friar also addressed the return-on-investment debate directly, dismissing reports of poor AI returns — "I find that, you know, spurious report" — and offered two examples: Cisco engineers saving 1,500 hours per month with OpenAI's Codex — "Cisco, their engineers are saving 1,500 hours per month" — and breakthroughs in 40 previously unresolved children's diseases at Boston Children's Hospital. She said OpenAI is holding back a next-generation model for additional safety and alignment work — "we want to do a few more turns on safety and alignment" — and the episode noted the company is delaying its planned IPO. All of the revenue and usage figures are the company's own; treat them as OpenAI's claims, not audited results. m:…7518

Security now, open-source pressure later

Cisco CEO Chuck Robbins, also on Bloomberg, argued the build-out cannot be throttled — "I don't think there's any slowing it down. It's an interesting philosophical discussion, but I don't think the reality is that it can be slowed down" — and said the priority should be faster cyber defenses. His warning: AI is "democratizing the adversary's capability" — "You don't have to be a rocket scientist or a computer scientist to be a very good hacker at this point." He named the Middle East and India as Cisco's two big growth regions outside the U.S.: "I think the Middle East is much like India. These are the two big areas that I see outside the U.S. where there's real growth opportunity for us." m:…5447

On the a16z show, Trillium founder Nathan Lambert argued the competitive dynamics cut the other way. Model progress comes down to three forms of capital, he said — "When I think about model progress, it comes down to different forms of capital, and that is compute data and talent" — and he pushed back on underestimating Chinese labs: "The Chinese are so good at building models. To say that China is anything but equal on talent, you're going to have a hard time." The episode's framing is that the open-source ecosystem is catching up through strategic model releases, which pressures the margins of the big spenders financing the build-out. Taken together, the week's finance podcasts tell one story in two halves: the credit market is starting to price the risk that AI demand disappoints, while AI's biggest private company insists the demand is already compounding. That tension is the thing to watch — the rest is commentary. m:…0913 m:…2186