Data centers are the true AI bottleneck
The guest argued that the primary constraint on AI compute growth over the next 15 to 18 months is physical data center space and power, rather than chip manufacturing capacity.
The argument
Despite building revolutionary wafer-scale chips, the guest noted that the entire industry is currently constrained by physical buildings, real estate, and power availability, which cannot be scaled as quickly as silicon fabrication.
The thesis, stress-tested
✓ What validates it
- ✓Utility companies approving large-scale power allocations for AI data centers
- ✓New data center capacity coming online faster than expected
▸ Risks discussed
- ▸Power grid capacity limitations in key data center hubs
- ▸Long lead times for data center construction and utility approvals
Hear it yourself
"That's how you put a processor, how you put a piece of silicon on a motherboard, deliver power and IO to it. We have to make inventions in power delivery. Right? When you build a giant chip, you're gonna deliver way more power to it than if you do a chip the size of a postage stamp. We had to invent ways to cool it. We had to write new types of software that ran on it. All of these had never been done before, and it was a a decade long process. It took us five years and about $500,000,000 to deliver the first one, and it's been an extraordinary run since. In December, we signed a a deal with OpenAI north of $20,000,000,000, one of the largest contracts ever signed in Silicon Valley."
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE