Dylan Patel — Deep dive on the 3 big bottlenecks to scaling AI compute

Money committed to AI infrastructure can precede usable computing capacity by years. SemiAnalysis CEO Dylan Patel explains in this interview that capital spending includes deposits for future power equipment and construction, alongside hardware entering service sooner. Conversely, some capacity delivered in a particular year reflects spending made earlier. That timing matters when comparing an investment announcement with a lab's ability to train models or serve customers: the announcement alone does not tell you when the machines will be available. His account frames computing supply as a sequence of commitments and deliveries whose schedules must line up.

“You sort of have this dynamic that is quite interesting in that an H100 is worth more today than it was three years ago.”

Dylan Patel

“Again, this is the most complicated machine that humans make, period, at any sort of volume.”

Dylan Patel

“You go down the supply chain, everyone’s doing X - 1. In some cases, they’re doing X ÷ 2, because they’re not AGI-pilled.”

Dylan Patel

“But in reality, he’s screwed the pooch compared to OpenAI, whose approach was, “Let’s just sign these crazy fucking deals.””

Dylan Patel

“It’s fast timelines, the US wins; long timelines, China wins.”

Dwarkesh Patel