$AVGO CEO Hock Tan just called out the real bottleneck in AI infrastructure — and it's not silicon.
Chip and memory supply? Easing. Capital to fund the actual build? That's the constraint now.
Data centers need more than GPUs. They need buildings, power infrastructure, and massive upfront investment. Even when chips are available, growth stalls if buyers can't raise the cash to deploy them.
Tan's take on AI investors is sharp: many are like casino gamblers — chasing high returns without understanding what they own.
Lots of companies are spending heavily on AI, but few are converting that spend into real revenue or profit. OpenAI's sales are climbing fast, yet model costs are so high that top-line growth alone doesn't prove the business model works.
This is the shift from "can we build it?" to "can we afford to build it — and will it pay off?"
The next phase of AI isn't about chip availability. It's about who has the balance sheet, the power contracts, and the business model to make the math work.
Chip and memory supply? Easing. Capital to fund the actual build? That's the constraint now.
Data centers need more than GPUs. They need buildings, power infrastructure, and massive upfront investment. Even when chips are available, growth stalls if buyers can't raise the cash to deploy them.
Tan's take on AI investors is sharp: many are like casino gamblers — chasing high returns without understanding what they own.
Lots of companies are spending heavily on AI, but few are converting that spend into real revenue or profit. OpenAI's sales are climbing fast, yet model costs are so high that top-line growth alone doesn't prove the business model works.
This is the shift from "can we build it?" to "can we afford to build it — and will it pay off?"
The next phase of AI isn't about chip availability. It's about who has the balance sheet, the power contracts, and the business model to make the math work.