Broadcom Hock Tan’s casino metaphor raises who will fund AI next

Tan’s casino comparison came on Monday, October 12, at the DBS Asia Future Forward Conference in Singapore. He was on a panel about where the next trillion dollars for AI will come from, alongside Intel CEO Lip-Bu Tan, with DBS chief executive Tan Su Shan moderating.

His core claim was that many investors set sky-high expectations, then discover the outcome is far less certain than they assumed. He was careful to say “a lot of” investors, not all of them.

He went further on businesses. Most companies, he said, still haven’t found AI uses that clearly pay for themselves. He compared corporate AI experiments to a kid at an arcade who burns through $10 of tokens in 10 minutes and walks away having learned nothing. In AI, a token is the small chunk of text a model reads or writes, and companies pay for every one, so the arcade joke was a pointed one.

He also gave a mixed verdict on OpenAI. Its annualized revenue of about $50 billion was, in his words, “by itself, not that bad” for a startup. But he noted the company is still losing money because training new models costs so much, so the revenue figure says little yet about long-term profit.

Here is the warning that matters most. Both Tans on the panel said the next big limit on AI may be money itself, ahead of chips, memory or electricity. Hock Tan called capital possibly the single biggest constraint, and Lip-Bu Tan pointed to the cost of borrowing and to investors who, sooner or later, want a return.

That is a big claim because the AI building spree runs on outside cash. Data centers are funded with loans, private credit and shares sold against revenue that often doesn’t exist yet. If lenders and shareholders get nervous, new projects stop long before anyone runs short of chips.

There are early signs of that nervousness. In recent weeks smart-ring maker Oura, Nvidia-backed data center firm Firmus and SoftBank-backed SB Energy all postponed plans to list on the stock market. Higher bond yields are making borrowing pricier too.

Yet Tan doesn’t think demand is fading. He said the need for computing power still outstrips supply and could stay that way through 2029, perhaps into 2030, with wafer shortages (the silicon discs chips are cut from) easing around 2028 or 2029. Lip-Bu Tan said he saw no big slowdown in the next two to three years.

No crying in the casino

— Elon Musk (@elonmusk) October 12, 2026

Elon Musk, who runs xAI and is one of the biggest buyers of AI computing in the world, replied on X to a post about Tan’s comments. His answer was five words long: “No crying in the casino.”

Musk didn’t explain further. The line plays on the idea that anyone who chooses to bet shouldn’t complain about the odds, which reads as a swipe at both nervous investors and at Tan for scolding them. It drew wide attention mostly because it came from someone spending heavily on the same AI buildout Tan was talking about.

There is a twist in who delivered the warning. Broadcom is one of the biggest winners of the AI boom. Its AI chip revenue hit $16.7 billion last quarter, up 221%, and the forecast for this quarter is $21.7 billion. For the full fiscal year Tan is looking for roughly $58 billion, and he has pointed to $230 billion by fiscal 2028.

Broadcom is also borrowing $60 billion so that Anthropic and other buyers can afford its chips and the data centers around them. In other words, the company is lending a hand to the same financing chain Tan says is under strain.

Both things can be true. Broadcom builds custom chips for a small group of very large cloud customers under deals signed years ahead, so it sits further from the riskiest bets. His warning was aimed mostly at the companies and investors downstream of those orders.

Markets reacted fast. Chip stocks fell across Asia the next day, with South Korea’s Kospi index dropping more than 7% during trading, Samsung Electronics down 9% and SK Hynix down 10% by midday. Broadcom’s own shares slipped 1.88% to $354.73, about 28% below their 52-week high.

For everyday people, the effects are slower but real:

• Your savings: if you hold an index fund or a retirement plan heavy in tech, AI chip stocks now make up a large slice of it, so mood swings like this show up in your balance.
• Your apps: tighter funding tends to push companies toward AI features that earn money now, like customer service or document tools, and away from expensive experiments.
• Your bills: AI firms under pressure to show profit may lean harder on paid tiers and price rises rather than free access.

None of this means AI is about to stall. Morgan Stanley researchers expect AI capital spending of about $800 billion this year, rising to $1.1 trillion in 2027. The open question is who keeps paying, and on what terms.