scripod.com
Etched - Building AI Hardware to Make Inference Faster and Cheaper - [Invest Like the Best, EP.480]

Overview

Shownote

Highlights

Transcript

Chapters

Pins

Etched - Building AI Hardware to Make Inference Faster and Cheaper - [Invest Like the Best, EP.480]

Invest Like the Best with Patrick O'Shaughnessy

Jun 30
Invest Like the Best with Patrick O'Shaughnessy

Invest Like the Best with Patrick O'Shaughnessy

Jun 30
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

My guests today are Gavin Uberti and Rob Wachen, the founders of Etched.  A few years ago, when they set out to build a better AI chip than the largest companies in the world, almost everyone I called told me it could not be done. They have since done it,...

Highlights

In this episode, the founders of Etched, Gavin Uberti and Rob Wachen, discuss their journey to build a specialized AI inference chip that challenges industry giants like NVIDIA. They detail the technical bets behind their architecture, their unconventional hiring strategy, and their vision for the future of AI infrastructure.
00:00
Welcome to Invest Like the Best.
03:54
First hardware company founded after ChatGPT to tape out a working chip.
03:58
General-purpose design creates inefficiencies.
10:22
They developed a low-voltage power delivery method to avoid thermal throttling
14:17
Improving throughput and concurrency is critical to scaling AI for billions of users.
20:29
A cancer diagnosis gave him a drive to do something impactful.
31:08
Contrarian candidates self-select for risky startups
36:31
Flood the binding constraint.
38:02
Speed is critical in AI inference.
44:00
Concurrency is the key metric for AI hardware
55:15
Winning is existential for us.
57:47
Many experiments fail, but only a few need to succeed.
1:05:41
We raised the largest AI chip Series A ever
1:08:43
Variant perception creates opportunity.
1:15:48
Adding our hardware alongside GPUs is positive-sum, not zero-sum.
1:16:42
Math is cheap, memory is expensive.
1:20:05
Inference will become a majority of global GDP
1:23:44
Trillion-dollar data centers are inevitable.
1:26:23
Risking paralysis for a higher chance of survival.

Chapters

Welcome to Invest Like The Best
00:00
Gavin Uberti and Rob Wachen
02:07
Two 21-Year-Olds Taking on NVIDIA
03:54
The Two Technical Bets Behind Their Architecture
07:52
Why Inference Becomes the Biggest Market
14:15
Rob and Gavin's Origins Stories
20:23
How They Recruit Industry Legends
28:38
Moving a Dozen Engineers to Bangalore for Six Months
36:30
Speed Wins
38:01
Getting More Concurrency Out of Every Megawatt
43:58
Vertical Integration
52:44
Hardest Obstacles to Overcome
57:43
Raising The Largest AI Chip Series A Ever
1:01:09
TSMC
1:06:29
Designing Gen 2 for Gigawatt-Scale Production
1:13:20
Why Machines Don't Think Like People
1:16:42
A Year of Compute Compressed Into a Month
1:20:03
The Trillion-Dollar Data Center
1:23:44
The Kindest Thing
1:26:19

Transcript

Speaker 1: I know firsthand how complex the tech stack is for asset managers. And seemingly every new tool and data source makes the problem even worse, adding more complexity, more headcount, and more risk. Ridgeline offers a better way forward, one unifi...