Google TPU
Google's custom-designed AI accelerator chips for machine learning workloads
Recently discussed in
What creators say
"We love TPUs, we're going to make TPUs work. Yes, we love Trainium, we're going to make Trainium work. Every chip has a comparative advantage; we have to find that advantage and then squeeze it in that direction."
More from Neil Movva →"We came up with TPUs, which are very specialized for essentially low precision dense linear algebra, which is at the heart of nearly all of the modern machine learning algorithms we use today. That system produced a chip a couple years later that ..."
More from Jeff Dean →"It fundamentally is not existential for my company for this product to win. For Google with TPUs, their revenue comes from search. Google won't fail if TPUs fail."
More from Gavin Uberti →"If Google has TPUs, Amazon have Trainium, while OpenAI now has Jalapeno."
More from Harry Stebbings →"Those TPUs are I mean that that's the lynch pin. And if you know they're 2x the power efficiency, but they're they're in the hunt now competing vertically."
More from Peter Diamandis →"You look at today where the latest Google TPU and where the Nvidia chipsets are at we're talking five petaflops"
More from Mario Harik →"Those decisions made 10 years ago set up the TPU as the perfect inference engine for the reasoning tech steps I'm describing."
More from Eric Schmidt →"the TPU was developed by the TPU team. DeepMind didn't realize how important it was going to be, how compute constrained they were going to be. So the TPU went and sold it to Anthropic"
More from TBPN →"I don't think people realize that Alexa and Google Home and Apple Home are about to use this for real."
More from Gary Vaynerchuk →"Anthropic signed a deal for a million TPUs"
More from 20VC with Harry Stebbings →"Largely because the margin on NVIDIA chips is insane, and Google can develop everything from top to bottom to fit their stack and not have to pay this margin."
More from Nathan Lambert →Created by
"I created Google TPU and at Google there was this time where AlphaGo running on GPUs had an ELO score of about 3,200. When put on TPUs, it actually went to like 3,900 — it jumped dramatically. The exact same model. What changed was the ability to ..."
More from Jonathan Ross →Keep exploring
About this data
- ✓44,000+ mentions extracted from real podcast transcripts
- ✓Every quote links to the source episode
- ✓Sponsored mentions are always labeled
- ✓Critical takes are shown, never filtered out