Broadcom CEO Hock Tan reaffirms AI revenue targets amid market concerns

Wall Street spent the week panicking about an AI slowdown. Broadcom’s CEO essentially told everyone to calm down.

Hock Tan appeared on CNBC’s “Mad Money” on September 14 and confirmed that the chipmaker’s AI semiconductor revenue guidance hasn’t budged. The targets: roughly $115 billion for fiscal 2027 and $230 billion for fiscal 2028. When asked if recent calls from AI leaders to pump the brakes on development had changed his outlook, Tan’s answer was blunt.

“No, not in the least.”

The reassurance came after Broadcom shares dropped 4.8%, swept up in a broader AI infrastructure selloff. That selloff was triggered by an open letter from Anthropic CEO Dario Amodei arguing for a more measured approach to AI development, a letter that drew support from Sam Altman and Elon Musk, among others.

The numbers behind the confidence

Tan’s conviction isn’t coming from thin air. Broadcom reported Q3 2026 AI semiconductor revenue of $16.7 billion, representing 221% year-over-year growth. The company also raised its full-year fiscal 2026 AI semiconductor guidance to $58 billion, up from a prior target of $56 billion. So while the rest of the market was digesting cautionary rhetoric from AI executives, Broadcom was quietly bumping its numbers higher.

The growth engine here isn’t just AI model training. Tan specifically highlighted inference demand as a critical driver. Inference is what happens after a model is trained: every time you ask ChatGPT a question or an autonomous vehicle processes sensor data, that’s inference compute at work. The hardware needs are scaling fast.

Anthropic: from cautionary voice to top customer

There’s an irony worth noting. The same Anthropic whose CEO’s letter spooked the market is poised to become Broadcom’s largest custom-chip customer by fiscal 2027, according to Tan’s projections.

Broadcom has been aggressively positioning itself as the go-to supplier of custom AI accelerators and high-speed networking components for hyperscalers. Unlike Nvidia, which dominates with general-purpose GPUs, Broadcom’s play centers on application-specific integrated circuits, or ASICs, custom-designed chips tailored to a particular customer’s workload. For companies like Anthropic, Google, and Meta that run inference at planetary scale, custom silicon can deliver meaningful efficiency gains over off-the-shelf alternatives.

What the selloff actually signals

The 4.8% drop in Broadcom shares, alongside broader declines in AI infrastructure names, reveals something about market psychology more than fundamentals. The Amodei letter argued for developing AI more carefully. Markets briefly priced in fewer chip purchases. Tan’s appearance was designed to close that gap.

For Broadcom specifically, the risk calculus is straightforward. If inference demand continues scaling at anything close to current rates, the company’s targets look achievable. The counterargument is concentration risk. If Anthropic is projected to be Broadcom’s single largest custom-chip customer by fiscal 2027, any stumble at Anthropic creates outsized exposure. Broadcom’s fiscal 2028 target of $230 billion assumes that multiple hyperscaler relationships continue expanding simultaneously.

Investors watching this space should pay less attention to open letters about AI philosophy and more attention to quarterly capex disclosures from the major cloud providers. Those numbers, not sentiment, will determine whether Broadcom’s doubling-and-doubling-again thesis holds up.