broadcom Chief Executive Officer Hock Tan on Monday dismissed concerns about what a potential slowdown in frontier AI model development could mean for the chipmaker’s business, telling CNBC that the company is sticking to its long-term revenue goals.
Tan’s comments came as investors were spooked by an essay published over the weekend by Anthropic CEO Dario Amodei arguing for a moderation in the pace of model development. Amodei’s message, which was supported by OpenAI CEO Sam Altman and Elon Musk, urged investors in AI infrastructure providers to reconsider their computing demand expectations during Monday’s trading.
Shares of Broadcom, whose custom chip business counts Anthropic as one of its most important customers, fell 4.8%. The iShares Semiconductor ETF, a broad basket of semiconductor stocks, fell 5.6%. Stock prices of companies that sell other data center components also took a big hit.
“No, at least not,” Tan said on “Mad Money” when Jim Cramer asked if the AI slowdown discussion had caused him to reconsider Broadcom’s AI semiconductor forecasts for 2027 and 2028. “We think the demand for computing infrastructure, AI development or AI frontier models, and inference products that they deliver to the world will continue to be very strong and very durable.”
During Broadcom’s fiscal 2026 third-quarter earnings conference on September 2, Tan predicted that AI semiconductor sales would reach $115 billion in fiscal year 2027 and double again to $230 billion in fiscal year 2028. A better-than-expected 2028 target was one of the bright spots in Broadcom’s earnings report. Broadcom’s AI revenue consists of both custom AI accelerators and networking chips used in AI systems.
Tan also said that Anthropic will become Broadcom’s largest custom chip customer in 2027 and will maintain that designation in 2028, showing why Broadcom investors are so sensitive to Amodei’s comments. Google has historically been considered Broadcom’s largest custom customer, co-designing Google’s tensor processing unit.
In an interview with CNBC on Monday, Tan was particularly optimistic about the demand for inference, or the everyday use of AI models after training.
“I don’t know about training, but I find that inference continues to be very powerful if you want to productize it,” Tan said.
Amodei’s essay over the weekend intensified the debate over whether the AI industry is moving too quickly to develop powerful models. Amodei proposed a three-step plan to slow the pace of development without “sacrificing commercial advantage or America’s lead in AI.”
Tan said he agrees with Amodei about the need for certain limits on AI, but suggested he is less wary of where the technology is headed.
“As with any tool, it is important to have governance and safeguards in place for how the tool is used,” Tan told Kramer.
However, Tan insisted, “It’s not a living animal that goes wild on its own.”
Tan instead emphasized AI’s potential to improve productivity. “AI, generative AI, the creation of those frontier models… will create huge value,” Tan said, likening it to the industrial revolution that began in Britain in the 18th century.
“At the end of the day, it is still a tool to help our society, humanity, reach a better standard of living,” Tan said.
