Three years ago, two Harvard dropouts set out to build a better AI chip than the largest companies in the world. Almost everyone I called at the time said it was impossible. Today, Etched (@Etched) comes out of stealth with $800M total raised, $1B in signed customer contracts, and a working next-gen AI chip. This was my excuse to ask the two founders, @UbertiGavin and @robertwachen, every question I have about compute and inference. We discuss: - Why they built an entire rack and not just a chip - The two technical bets behind their architecture no one else has tried - How two founders in their twenties recruited industry legends - The night they nearly ran out of money - Why whoever produces the most tokens wins If you care about the future of compute, Gavin and Rob are two people to know. I think you will find the story of what they have built hard to forget. Enjoy! TIMESTAMPS 0:00 Intro 1:00 Why Nobody Believed Etched Would Work 14:06 Why Inference Is the Bottleneck 22:27 Gavin and Rob’s Origin Stories 33:24 Taking Huge Risks to Move Faster 49:43 Kernels, Compilers, and the AI Stack 1:02:08 Raising $100M to Survive 1:16:00 The Future of Models, Agents, and Intelligence

Jun 30, 2026 · 3:00 PM UTC

66
219
2,351
1,464,999
Sort replies: Relevant Recent Liked
Incredibile team/story...
15
5,237
What an amazing interview. The depth of understanding, vision, hard questions asked, challenges confronted, a great achievement with the first product, and dazzling, almost dizzying, future.
7
1,049
The age of impossibility is over. Congratulations 🍾
8
1,339
Rebels, backed by ribbita-by-virtuals:native (spelled backwards) Capital 🐸
5
874
can you comment on this so i can stop getting run over on my microscopic nvda short @citrini
6
3,570
Contrarian and Right = Big Returns!
4
1,271
Was a fun one @patrick_oshag :)
4
1,030
guy on the far right has decabillionaire physiognomy, im in
5
1,081
So very cool
2
1,038
Gribbit 🐸
2
768
It’s a sight to see genius at the point of thought formulation. Great Q&A.
2
809
CUDA is the moat China will just outbuild and out-innovate this startup USA Ai will be just Gov-protected
1
1
1
909
Bold vision creates industry-defining breakthroughs.
1
424
真牛逼胆大敢干的人都没饭吃不敢靠头这么勇
373
@domcooke Does the OpenAI/Broadcom chip come up or was it recorded prior?
1
3,891
I reached out to @UbertiGavin after I heard him on @InvestLikeBest in 2023. It was abundantly clear that he's a true outlier talent, and talent magnet, and we made Earthshot's largest initial investment up to that point. Thanks for helping me source a banger @patrick_oshag!
1
14
4,591
inference is literally the battleground now and most people still don't get it
5
2,972
This is unbelievable achievement. At this age I was just trying to finger everyone at Infernos
6
1,803
@nikitabier can we have a watch later option like YouTube in x
2
650
People are acting like Etched is bad for $MU or about to dethrone $NVDA. Neither makes much sense. Etched still relies on high-bandwidth memory—it doesn’t replace it. If its chips are deployed at scale, that’s potentially more demand for HBM, which benefits $MU. As for NVIDIA, building a fast chip is only a small part of the equation. NVIDIA’s moat is its entire ecosystem: CUDA, networking, software, developer tools, enterprise support, manufacturing scale, and years of optimization. That’s why customers stay. A startup can build an impressive ASIC. Replacing NVIDIA’s AI platform is an entirely different challenge. Bottom line: Etched isn’t a bearish thesis for $MU, and it’s nowhere close to making a meaningful dent in $NVDA today. Eventually will be part of bigger semis when it acquired !!
1
578
Excellent interview. Thanks!
1
621
betting against the "that's impossible" crowd is basically the entire history of hardware. congrats to the etched team, that's a wild run from stealth.
1
2,048
@etched and @patrick_oshag Thanks for the opportunity to take a peak into the future 🙏
1
727
impossible" sure changes meaning fast when you have $800M lol, curious what those people say now
1
2,703
Computers look again like in the 40s.. Make Computers Great Again. Everything is a cycle.
1
567
What an impressive tandem. Both seem like top shelf people. Excited to see what they do!
253
this is history
1,757
Brilliant episode
50
Perfect founder duo.
395
Inference is eating the world and the companies building dedicated silicon for it will own the next decade.
35
By the time they bring their tech to market, that too will become outdated. This is the age old danger hard tech = fast obsolescence.
347
Great show. I learned a lot. Thx. 😊 Now I have to start thinking about how to get access to run 1QT tokens through. Parallels at 10x is fascinating. I've been thinking out how to build a JARVIS. But that sent me down a path to create a new ways to prompt, route, research, all w/swarms. And now to step distillation to retraining autonomously each each expert simultaneously w/tier router. To simplify it's a path to taking a giant high intelligence open source open weights LLM Intelligence Index (ii) number of 50 and end up with a <10B S LM w/ii above 90. This is the best way to get to local genius inference. And w/Etched computational advancements the world changes radically immediately, imo. Anyways good interview.
1,933