Broadcom CEO Hock Tan Compares Many AI Investors to Casino Gamblers
Broadcom expects AI revenue to hit $230 billion by 2028. Its own CEO says plenty of the money behind that number belongs in a casino, not a portfolio.
Hock Tan doesn't usually undercut the story that has tripled Broadcom's AI business in two years. But speaking in Singapore on Monday, the Broadcom CEO told investors flatly that a lot of people pouring money into artificial intelligence are behaving like gamblers, not investors. "A lot of investors have no clue what they are investing in," Tan said, adding that they set "very extreme expectations only to find out that it's almost like they are in a casino," according to Bloomberg.
Tan aimed the comment squarely at the money chasing frontier large language models, the OpenAIs and Anthropics of the world, where he argued enthusiasm has outrun any real grasp of the underlying economics. He didn't stop at investors. He took a swing at how enterprises are spending on AI too, comparing their token budgets to a kid blowing through arcade credits. "Your parents give you $10 of tokens, you finish it in 10 minutes and you haven't learned anything," Tan said. "That's what's happening in enterprises today."
The line lands with extra weight because of who is saying it. Broadcom is not a bystander worried about an AI bubble popping somewhere else. It is one of the biggest direct beneficiaries of the spending Tan is now calling reckless. The company's AI semiconductor revenue hit $16.7 billion in the third quarter of fiscal 2026 alone, more than tripling year over year, and Broadcom raised its full-year AI revenue guidance to $58 billion, up 186% from the prior year. Management has told investors to expect that number to double to $115 billion in fiscal 2027, then double again to $230 billion in fiscal 2028. Those forecasts rest on custom AI chip deals with Google, OpenAI, Anthropic and Meta, the exact customer list Tan's casino comment implicitly puts under a cloud.
So Tan isn't calling the whole industry a mirage. He's drawing a line between the infrastructure layer, where real chips ship and real contracts get signed years out, and the speculative layer sitting on top of it, where capital flows into frontier model developers on the strength of a growth story rather than a balance sheet. That's a more precise claim than the generic "AI bubble" chatter that's dominated markets this fall, and it's harder to wave off coming from someone who sells the shovels.
OpenAI paid Microsoft 17.2 billion dollars in 2025 alone, more than its entire annual revenue, while its enterprise unit grew 107% and still wasn't enough to hit the revenue figure investors expected. - OpenAI's leaked financial losses raising investor doubts - AI bubble evidence from OpenAI's revenue numbers
OpenAI's annualized revenue reached about $50 billion at the end of September, a milestone that shows how fast demand for frontier AI models is scaling even as the economics remain unsettled. The company is still private and continues to spend heavily on training and infrastructure, making durable profitability a separate and much harder problem than top-line growth.
Tan's remarks arrive in the middle of a broader wave of insider hedging that's rattled AI-exposed stocks this fall, from warnings around memory chip valuations to rising anxiety in credit markets over how AI infrastructure debt is being financed. What makes Tan's comment different is the source: Broadcom's stock price is directly levered to AI capex, and its multiyear semiconductor outlook depends on the buildout continuing. When the CEO most dependent on that demand says many investors do not understand what they own, that's not hedging. That's an insider naming the risk in his own backyard.
That distinction should worry chipmakers less than investors. Broadcom has told investors it has secured the supply needed to support its fiscal 2027 and fiscal 2028 AI semiconductor outlook, and Tan has said demand already exceeds the company's fiscal 2027 forecast. The part he is nervous about is upstream, in the public markets pricing AI stocks as if every dollar of capex converts cleanly into durable earnings. He didn't offer a fix. He just said the bet is bigger than most people placing it understand.
Nvidia Pushes RTX Video Upscaling Into ComfyUI's Open-Source Pipeline • How Alexandr Wang Pushed Meta Muse to No. 1 in the App Store, Ahead of ChatGPT • Linus Torvalds used Claude Code to rewrite how Linux reads cached files
This article is posted in AI News, check it out for more related stories.
Broadcom’s AI chip revenue is still growing rapidly, but investors punished the stock after guidance failed to clear the market’s high bar. The reaction shows how custom silicon is becoming a harder story to price as hyperscaler spending moves from promise to delivery. - why AI chip stocks fell despite strong growth - AI semiconductor revenue growth expectations for investors
