As debate continues to intensify across the technology sector regarding whether frontier AI model development should be tempered, Broadcom (AVGO.US) Chief Executive Hock Tan has joined the chorus pushing back against those concerns. Following NVIDIA (NVDA.US) CEO Jensen Huang's recent public opposition to the so-called "AI brake" thesis, Tan is now seeking to reassure investors about the sustainability of his company's growth trajectory.
In a recent interview, Tan directly addressed worries that a potential slowdown in frontier AI model development could adversely impact the chipmaker's business, reaffirming that the company remains firmly committed to its long-term revenue goals. His statements come at a time when market sentiment has been rattled by an essay published over the weekend by Anthropic CEO Dario Amodei, which advocated for a more measured pace of model development. Amodei's perspective has also garnered support from OpenAI CEO Sam Altman and Elon Musk, prompting investors in AI infrastructure providers to reassess their expectations for computational power demand during Monday's trading session.
The market reaction was swift and significant, with Broadcom shares declining 4.8%—a notable move given that its custom chip business counts Anthropic among its most important clients. The broader semiconductor landscape felt the impact as well, with the iShares Semiconductor ETF (SOXX.US) falling 5.6%, while companies selling other data center components also experienced substantial losses.
Addressing the ongoing debate, Huang had earlier dismissed predictions that AI could pose an existential threat to humanity during a discussion at the All-In Summit in Los Angeles on Monday, characterizing such claims as lacking scientific foundation. Just a week prior, Huang had made a characteristically pointed observation, suggesting that panic surrounding AI safety might simply be generating commercial opportunities for the cybersecurity industry—noting that "what better way to create demand than to create a problem?"
When prominent financial commentator Jim Cramer asked during a program whether any elements of the AI slowdown debate had caused Tan to reconsider Broadcom's AI semiconductor projections for fiscal years 2027 and 2028, the CEO's response was unequivocal: "No, not at all." Tan sought to calm market anxieties from both demand and revenue perspectives, stating, "We see that whether it's the computational infrastructure for frontier model research and development, or the AI inference computing power for deployed products, demand remains exceptionally strong—and I believe this demand has substantial staying power."
Tan's confidence echoes his previous guidance from the company's fiscal third-quarter earnings call on September 2, during which he projected Broadcom's AI semiconductor revenue would reach $115 billion in fiscal 2027, with expectations of doubling to $230 billion in fiscal 2028. This ambitious 2028 target, which surpassed market expectations, emerged as a highlight of that earnings report. Broadcom's AI revenue comprises both custom AI accelerators and networking chips designed for AI systems.
Tan further revealed that Anthropic is positioned to become Broadcom's largest custom chip customer by 2027 and is expected to maintain that position into 2028—clarifying why investors in the company respond so sensitively to Amodei's public statements. Prior to this shift, Google (GOOGL.US) had held the distinction of being Broadcom's largest custom chip client, with the two companies having collaboratively designed Google's Tensor Processing Units (TPUs).
During Monday's interview, Tan expressed particular optimism regarding inference computational demand, which refers to the processing power required during day-to-day operations after AI models have completed their training phase. "I can't speak definitively about training, but once models become commercialized products and enter inference scenarios, I see demand that will continue to grow extremely strongly," Tan remarked.
The ongoing discourse was further intensified by Amodei's weekend essay, which escalated the industry-wide debate about whether AI development of powerful models is progressing too rapidly. Amodei proposed a three-step plan designed to moderate the pace of development while presumably maintaining "commercial advantage and America's leadership position in AI." Tan indicated agreement with Amodei's general sentiment that some form of AI regulation is necessary, though he suggested his own concerns about the direction of technological progress are less pronounced.
"Like any tool, it's important that we have governance and safeguards around how we use it," Tan stated. However, he offered his own counterpoint, asserting that "AI is not a living organism that will run wild and uncontrollably." Tan pivoted instead to emphasize AI's potential for enhancing productivity. "AI, generative AI, the deployment of those frontier models... will create tremendous value," he said, drawing a parallel to the Industrial Revolution that began in 18th-century England. "Ultimately, it remains a tool that will lift our society and humanity to a higher standard of living."