Broadcom CEO addresses Anthropic's slowdown push, says AI revenue targets haven't changed

Broadcom CEO Hock Tan used an appearance on CNBC's "Mad Money" on Monday to push back on fears that a slowdown in frontier AI development would dent the chipmaker's business, saying the company stands by its long-term revenue targets.

Asked by host Jim Cramer whether the AI slowdown debate had made him reconsider Broadcom's fiscal 2027 and 2028 AI semiconductor forecasts, Tan answered, "No, not in the least." He added that demand for compute infrastructure — for AI development and frontier models, and for inference on the products they power — "continues to be very strong and, I believe, very durable."

The debate was set off by a weekend essay from Anthropic CEO Dario Amodei, who proposed a three-step plan to moderate the pace of model development without "sacrificing commercial advantage or the United States' lead in AI." OpenAI's Sam Altman and Elon Musk have voiced support, and AI infrastructure stocks sold off during Monday's session.

Amodei's essay hit Broadcom directly because of its customer mix: Anthropic is one of the most important customers of Broadcom's custom-chip business. Tan has said Anthropic is on track to become Broadcom's largest custom chip customer in 2027 and to keep that status in 2028; the slot has historically belonged to Google, with which Broadcom co-designs TPUs.

The market reaction was immediate: Broadcom shares fell 4.8% on the day, the iShares Semiconductor ETF dropped 5.6%, and other data-center component makers were hit hard as well.

On its fiscal 2026 third-quarter earnings call on Sept. 2, Broadcom guided to $115 billion in AI semiconductor revenue in fiscal 2027, doubling to $230 billion in fiscal 2028 — a target that beat expectations and stood out as a highlight of the report. Broadcom's AI revenue combines custom AI accelerators and the networking chips used in AI systems.

Tan said he agrees AI needs some restrictions, but suggested he is less alarmed than Amodei: "Like any tool, it's important to put governances, safeguards on how we use the tool," he said. "It's not a live animal that will run wild by itself." He emphasized productivity instead, likening generative AI and frontier models to the Industrial Revolution that began in 18th-century England, and singled out inference demand: "I don't know about training, but when you want to productize inference, I see it continuing to be very, very strong."

What to watch: whether Broadcom can deliver on its fiscal 2027 and 2028 AI revenue targets, which depend heavily on its largest customers' capital spending, and whether Amodei's slowdown push changes hyperscaler procurement plans in the coming quarters.