Broadcom CEO rejects AI slowdown concerns and reiterates the company's $115 billion annual revenue target.

Broadcom CEO Hock Tan publicly downplayed market concerns on Monday that a slowdown in AI model development could dent chip demand, emphasizing that the company’s AI semiconductor revenue forecasts for fiscal 2027 and 2028 remain unchanged. The remarks followed a weekend post by Anthropic CEO Dario Amodei calling for a pause in frontier AI development, which triggered a broad selloff across AI infrastructure stocks. Amodei’s proposal drew public support from OpenAI CEO Sam Altman and Tesla CEO Elon Musk, prompting investors to reassess the outlook for compute demand. Broadcom shares fell 4.8% on the day, while the iShares Semiconductor ETF tumbled 5.6%, with other data center component suppliers also trading lower.

During an interview with CNBC’s “Mad Money” host Jim Cramer, Tan was asked whether the AI slowdown debate had prompted him to reconsider Broadcom’s AI chip revenue projections for fiscal 2027 and 2028. He responded emphatically: “No, not at all.” He elaborated that demand for compute infrastructure—whether for AI research and development, frontier AI models, or the inference compute needed to bring products to market globally—will remain robust, adding that he believes this demand has strong staying power. During Broadcom’s fiscal 2026 third-quarter earnings call on September 2, Tan projected that the company’s AI chip business would generate $115 billion in revenue in fiscal 2027 and double to $230 billion in fiscal 2028. That fiscal 2028 target, which exceeded market expectations, was a key highlight of the report. Broadcom’s AI revenue streams include custom AI accelerators and networking chips used in AI systems.

Tan also revealed that Anthropic is on track to become Broadcom’s largest custom chip customer in 2027, a position it will maintain through 2028. This shift explains why Broadcom investors are particularly sensitive to Amodei’s remarks, as Google has long been viewed as the company’s largest custom chip client following its co-development of Tensor Processing Units (TPUs). Tan expressed particular optimism regarding “inference” demand—the everyday computational load generated after models complete training. “I can’t speak to the training side, but once you productize inference capability, I believe the related demand will remain very, very strong,” he said.

Amodei’s weekend essay intensified the debate over whether the AI industry is advancing too quickly, proposing a three-step plan to slow development without sacrificing commercial advantage or America’s leadership position in AI. Tan agreed that certain guardrails on AI are necessary but remains unconcerned about the technology’s trajectory. “Like all tools, it’s important to establish governance rules and safety measures around how it’s used,” Tan said. However, he stressed: “It is not a living thing that will spiral out of control on its own and wreak havoc.” Instead, he emphasized AI’s potential to boost productivity, drawing a parallel to the Industrial Revolution that began in 18th-century Britain. “Artificial intelligence, generative AI, and the emergence of these frontier models… will create enormous value,” Tan said. “At the end of the day, it’s just a tool that can propel our society and humanity toward a higher standard of living.”

The AI slowdown debate unfolded against the backdrop of Anthropic’s impressive financial results disclosed ahead of its initial public offering. Preliminary figures show the company generated more than $11.5 billion in revenue in the second quarter of this year, marking a nearly 15-fold surge from $787 million in the same period of 2025 and surpassing its internal forecast of $10.9 billion. Adjusted operating profit for the quarter reached $559 million, representing the company’s first profit under that metric. Anthropic has selected the Nasdaq as its listing venue, with the offering potentially ranking among the largest IPOs in history.

Market analysts offered differing interpretations of the motivations behind the slowdown calls. Some argue that AI giants collectively advocating for a deceleration on the eve of an IPO may reflect multiple considerations: highlighting the technology’s dangers to underscore its value, cementing market position by participating in safety standard-setting, and establishing legal defenses against potential safety incidents post-listing. Aiden Gomez, CEO of Canadian AI firm Cohere, publicly criticized the move, calling it “a wolf in sheep’s clothing” and “a cartel by another name” for a handful of dominant market players to set global rules and safety standards. Conversely, some investment experts believe the market reaction may be overblown. Kim Jang-ryeol, head of the asset management center at South Korean asset manager Unistory, noted that it is still too early to judge whether an AI development slowdown will actually reduce infrastructure investment, arguing that maintaining both the pace of AI development and safety enhancements may paradoxically require more capital. “Whether AI growth decelerates will ultimately be decided by big tech companies, so watching their stock price trends is sufficient,” Kim said.

Reflecting the broader uncertainty, all three major U.S. stock indexes closed lower on Monday. The Dow Jones Industrial Average fell 152.01 points, or 0.29%, to close at 52,421.28. The S&P 500 dropped 37.03 points, or 0.48%, to 7,619.95. The Nasdaq Composite declined 146.62 points, or 0.56%, to 26,186.41. Among chip stocks, South Korea’s SK Hynix fell 7.6%, ASML dropped 7.2%, and Intel and Micron Technology each lost more than 5%. While Broadcom’s share price decline was notable, Tan’s firm stance provided investors with a measure of confidence. His optimistic assessment of inference demand aligns with market expectations for continued expansion in AI application layers. As Anthropic’s IPO progresses and discussions around AI safety standards deepen, the long-term demand outlook for the chip industry will remain a central focus for investors.