“The target hasn’t changed.” With three words, Broadcom Chief Executive Hock Tan moved to cool a market panic on September 21, pushing back against a fear that has gripped AI investors for days: that the spending boom on which his company’s growth rests is beginning to fade.
Tan was answering the question that has rattled chip stocks. After a weekend of public statements by AI executives calling for a slower, more cautious pace of development, investors sold AI infrastructure names and began to question whether the orders already booked would hold. Broadcom, which sells the custom chips and networking gear that hyperscalers use to build AI systems, sat directly in the line of fire.
The numbers behind the question are enormous. Analysts have put Broadcom’s AI-related revenue at $230 billion by 2028, a figure built on orders for the custom accelerators the company designs for its largest customers. Those customers, the cloud providers and AI labs that buy chips by the data center, are the same ones whose executives spent the weekend urging restraint. If they pull back, the arithmetic falls apart.
Tan’s answer was that the arithmetic still holds. The demand for compute infrastructure, he argued, remains strong, and the custom chip orders that anchor the forecast have not moved. In his telling, the selloff was a matter of sentiment, not of orders, and the company’s guidance stood exactly where it had been left on its last earnings call.
The distinction matters. Broadcom’s stock had already pulled back before Tan spoke, part of a broader retreat in semiconductor shares. Yet even as the share price fell, investors were adding to long positions rather than closing them, a signal that the pullback was being read as a correction in valuation, not a verdict on demand. The buyers, in other words, saw a cheaper entry into a story they still believed.
Broadcom’s story is unusual in the chip industry. Unlike Nvidia, which sells finished processors, Broadcom designs custom silicon for a small number of giant customers and earns revenue through multi-year contracts that are largely committed before a single chip ships. That structure gives the company unusual visibility into future demand, and makes its forecasts a closely watched barometer for the entire AI buildout.
It is also why the slowdown debate lands so hard on Broadcom. If the custom chip buyers were to slow their plans, Broadcom would know early, and the $230 billion target would be the first number to break. The fact that Tan chose to defend it publicly, rather than let the guidance speak for itself, suggests he saw the market’s reaction as out of proportion to anything happening inside the company.
Tan has spent years positioning Broadcom at the center of the AI supply chain, expanding far beyond the networking business that once defined the company. The bet has paid off in the share price and in revenue that has more than doubled in a short period. It has also concentrated the company’s fortunes in the hands of a few customers whose spending decisions now move Broadcom’s stock more than its own quarterly results.
The wider chip market’s retreat has shown how fragile the AI trade has become. The same executives whose companies buy the chips are now publicly debating whether the technology should advance more slowly, and every sentence is parsed for its effect on demand. In that environment, a chief executive saying “the target hasn’t changed” is a statement of defiance as much as a financial update.
For now, the market is taking Tan at his word. Broadcom’s shares stabilized after his remarks, and the long positions that built during the retreat suggest investors are willing to distinguish between a stock that is expensive and a business that is broken. The two, in a panic, are easy to confuse.
Broadcom’s rise under Tan has been built on acquisitions as much as on chips. He assembled the company through a series of deals that stretched from semiconductor design into enterprise software, capped by the purchase of VMware, and that software arm now gives Broadcom a second revenue stream to cushion the chip cycle. The AI custom-silicon business, however, has become the engine of the stock, and the $230 billion figure Tan defends is the clearest statement of how large he expects that business to grow.
What happens next depends on orders, not on words. Broadcom will report again in the coming months, and the guidance it gives will show whether the $230 billion figure is still the base case. Until then, the company’s message is simple and unchanged: the demand is real, the contracts are signed, and the target stands where it always did. Tan has given the market a number to hold him to, and it is now his job to make the coming quarters match it.
