Intel / video

The Economics of the AI Super Cycle: OpenAI's Compute Strategy (with Sachin Katti)

Source: youtube.com
Author: Stanford Online
Date: Indexed Aug 16, 2026

Context

An 's response time depends on more than how quickly its model generates text. In this OpenAI compute discussion, the workload includes searching for information, running tools, creating a computing environment to test code, inspecting results and returning to the model for another decision.

The speaker argues that these different steps require a mix of processors and accelerators, matched to the work each performs. The intended payoff is a more responsive working experience: speeding up text generation alone leaves delays elsewhere in the loop. This is an infrastructure rationale, rather than a measured end-to-end speedup.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.

Key quotes

“our vision is we want to get to a world where the human is the bottleneck”
Sachin

“today apps are a crutch”
Sachin

“a six-month lead in intelligence is an enormous lead”
Sachin

“I'm long open AI. I'm voting with my feet, but keeping a side.”
Sachin

Read the source youtube.com