Broadcom's CEO says the slowdown call changed nothing about AI demand
The market spent Monday repricing the AI buildout after Dario Amodei's call to pace the frontier. The people who actually write the compute cheques spent the same day saying the number has not moved.
Broadcom CEO Hock Tan said the slowdown debate has not touched his AI revenue forecasts. Asked by Jim Cramer on "Mad Money" whether anything in the argument over pacing frontier models had made him reconsider the company's fiscal 2027 and 2028 AI semiconductor targets, Tan answered "No, not in the least," adding that demand for compute infrastructure "continues to be very strong and, I believe, very durable." On the September 2 earnings call he put AI semiconductor revenue at $115 billion in fiscal 2027, doubling again to $230 billion in fiscal 2028; on Monday he said Anthropic is on track to overtake Google as Broadcom's largest custom-chip customer in 2027 and hold that spot in 2028. None of that stopped the tape: Broadcom fell 4.8%, and the iShares Semiconductor ETF dropped 5.6%. The distinction Tan drew is the one worth reading — "I don't know about training, but when you want to productize inference, I see it continuing to be very, very strong." Pacing a frontier model costs a lab training compute. It does not obviously cost an inference supplier anything, and Broadcom sells into both. On the safety argument itself he agreed with Amodei only halfway: governance and safeguards belong on the tool, but "it's not a live animal that will run wild by itself."
CrowdStrike CEO George Kurtz argued that slowing development cannot remove a risk that has already shipped. "The genie's out of the bottle," he said on the same programme — frontier and open-weight models already in circulation "can already be dangerous," so the security industry has to protect what exists while the frontier labs decide their pace. His answer was runtime rather than regulation: instrument the agents, put guardrails around them while they operate, because "you need equivalent or better AI defenses to combat the AI agents" — while warning that too much regulation "is going to stifle innovation" against a US lead he called narrow. Investors priced the same logic from the opposite side: CrowdStrike closed nearly 14% higher at a record above $235, and Palo Alto Networks gained just over 13%, both now up about 100% for the year while chip names sold off. We covered the defender's other argument this afternoon — Cohere CEO calls AI models the most potent cyber weapon ever built.
A London startup called Jack & Jill raised a $40 million Series A to sell agents on both sides of the job interview. The company offers Jack, an AI agent that hunts jobs and applies for candidates, and Jill, its counterpart for employers, per Axios; the round follows $20 million raised in 2025. Recruiting is the white-collar function LLMs were most often described as replacing in the abstract — this is the first sizeable bet that candidates will accept an agent negotiating against another agent, at which point the interview stops being the filter and the two agents' prompts become it.
What to watch: whether the other compute suppliers repeat Tan's "no change" when they report, or quietly trim.
Two CEOs spent Monday telling you the buildout is fine and a third told you the risk is already loose. Which one are you believing this quarter? Tell us in the comments.
