The people building AI's infrastructure just accused the people funding it of the same reckless speculation crypto lived through in 2021.
The Summary
- Broadcom CEO Hock Tan told investors in Singapore that many AI backers "have no clue what they are investing in" and behave like they're "in a casino"
- DBS CEO Tan Su Shan says the market is repricing AI financing risks, signaling a shift away from pure hype-driven capital allocation
- The criticism mirrors what crypto heard for years, except now it's coming from inside the AI supply chain, not from skeptics outside it
The Signal
Broadcom makes the actual silicon that powers AI infrastructure. When its CEO says many investors funding the AI boom don't understand what they're buying, that's not academic concern. That's the hardware guy watching the money guy place bets on a game whose rules he doesn't know.
The casino comparison is deliberate. Crypto heard it constantly during its 2017 and 2021 runs. Regulators, traditional finance, skeptical technologists all used gambling metaphors to dismiss digital assets as speculative froth. Now Tan uses the same framing for AI investors whose expectations are so inflated they've detached from operational reality.
"Many AI investors have no clue what they are investing in. It's almost like they are in a casino."
The timing matters. DBS, one of Asia's largest banks, is simultaneously repricing AI financing risks. That's not random. Banks don't reprice risk because they're worried about innovation. They reprice when capital discipline breaks down and defaults become probable.
What's happening: AI infrastructure costs real money. Data centers, GPUs, power, cooling, bandwidth. The companies building models need continuous capital to scale. When investors expect casino returns but infrastructure requires patient, grinding capital deployment, something breaks. Either expectations crash or funding dries up.
Key differences from crypto's casino phase:
- AI has clear enterprise revenue. Models are already in production at Fortune 500 companies.
- The infrastructure is physical. You can touch a GPU cluster. You can't touch a token in a liquidity pool.
- The criticism is coming from suppliers, not outsiders. Broadcom wants AI to succeed. They sell the chips.
The market repricing that DBS describes isn't a collapse. It's a sorting mechanism. Weaker players who can't justify their burn rate or revenue trajectory will get cut off. The ones with actual unit economics will keep building. This is what crypto never got in 2021: a repricing before the bottom fell out.
The Implication
If you're funding AI companies, Tan's warning is a mirror. Do you know what you're investing in, or are you betting on narrative momentum? The hardware guys are telling you the music might stop. The banks are already pulling back chairs.
For crypto, this should feel familiar and instructive. The "casino" critique isn't wrong when capital floods in faster than understanding. What matters is whether the underlying infrastructure justifies the hype after the repricing. AI has revenue and enterprise adoption crypto mostly lacked in 2021. But it also has the same problem: too much hot money chasing too few people who actually know what they're building.