AV

Amin Vahdat

Things Amin Says on Podcasts

Where to Find Them

Amin Vahdat writes a16z Podcast . They have also been a guest on Training Data , This Week in Startups and Catalyst with Shayle Kann .

Recently: “Google's AI Infrastructure Chief, Amin Vahdat, on the Physics & Economics of Frontier AI” on Training Data (October 2026); “Inside Google’s massive AI capex (live)” on Catalyst with Shayle Kann (April 2026); “Building the Real-World Infrastructure for AI, with Google, Cisco & a16z” on a16z Podcast (October 2025); “Building the Real-World Infrastructure for AI, with Google, Cisco & a16z” on a16z Podcast (October 2025); “What's Next for AI Infrastructure with Amin Vahdat | AI Basics with Google Cloud” on This Week in Startups (May 2025).

What They Said

“Our seven and eight-year-old TPUs are still at 100% utilization.” — Amin Vahdat, Training Data

The host asked whether Google goes back and swaps out the chips in older data centers, and how long an AI chip really stays useful. Vahdat repeated a remark he said had drawn more attention than he expected when he first made it. TPUs are Google's own AI accelerator chips.

Training Data · 2026-10-06 Permalink → Listen →
Training Data Around 49:02 into the episode
Speaker 1

…But now we wind up having individual model endpoints, and there actually can be variations of models. So now we have to spread these models out across the world, also accounting for locality. So they will be smaller. They'll actually be less vertically integrated, probably, on the inference side.

Speaker 2

…ago was very different, vastly less efficient than the ones that are being made today. Yes. Do you actually go back and like swap out the chips in those old data centers? And I know this kind of relates to there's been an ongoing debate, I think, of like what is the actual useful life of a chip.

Speaker 1

Yeah. Yeah. So, you know, I've been on record of saying this in one of my more, I was surprised by how much pickup this had. But, you know, I said that our seven and eight-year-old TPUs are still at 100% utilization.

Speaker 2

Wow.

Speaker 1

I think I saw

Speaker 2

this. I didn't realize this from the case. Yeah,

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

Collections They Appear In