Invest Like the Best with Patrick O'Shaughnessy

The episode is packed with genuine technical substance - MFU percentages on real workloads, the Dennard scaling argument for low-voltage inference, PD disaggregation mechanics, and the prefetching/parallelization philosophy - but long narrative stretches (cancer story, fundraising saga, robotics) significantly dilute the per-minute insight rate for a B2B operator audience.

“on GPUs you often get somewhere between 20 and 50% depending on the workload. And actually you can provably not run at 100% because you have a thermal issue where as you increase the flop utilization, you have more transistors going on and off, you draw more power, and the chip will self regulate and actually lower its clock speed”

“There is another very famous AI chip company that took 10 months to go from getting their silicon back to having them running inference in Iraq. And this was like publicly announced to their investors. And it was a really big deal. We were able to do it in 40 days.”

Several genuinely contrarian moves stand out - using Bitcoin miners as proof that AI chips can run at much lower voltages, the explicit choice to be kernels-first rather than build a compiler (and that only HFT firms initially understood this), and the 'assume it is possible' inversion of the standard engineering skepticism reflex. The macro 'inference will be the biggest market' framing is less original.

“Bitcoin miners run at under a quarter of the voltage of GPUs. So this is obviously physically possible. The question is, are there issues with GPU architectures that make it unable to run at these voltages?”

“The only people that took us seriously were in High Frequency Trading. They all hate compilers too. They all write their own kernels. And we've had dozens of people from High Frequency Trading join the team because they saw this philosophy too.”

Both founders are genuine practitioners who have actually taped out a working chip on a first attempt, raised $800M, and have $1B+ in customer demand - remarkable credentials for founders in their mid-20s. Gavin's prior kernel-engineering work at Xnor (acquired by Apple) and October (acquired by Nvidia) and the caliber of the team they recruited (Brian Leyler who built Nvidia's HGX/DGX) lend significant credibility, though commercial-scale shipping is still ahead.

“My first job ever was at a company called Xnor where I did kernels development. I was 17... Exnor got bought by Apple for 200 million, did the same thing at October, they got bought by Nvidia for hundreds of millions of dollars.”

“Brian started the HTX and DGX team at Nvidia, which was a majority of Nvidia's revenue. Tens of Billions of dollars a quarter. And the other two guys ended up investing, by the way.”

Unusually concrete throughout: named latency figures (4,000 nanoseconds Blackwell chip-to-chip, 5x improvement), voltage claims (under half the voltage of any other AI chip), engineering tolerances (50 picoseconds clock alignment, 2 billion times a second), operational milestones (700 FPGAs, 40 days vs 10 months, $103M Series A assembled soft commit by soft commit), and named individuals with verifiable roles.

“we had to go line up two clock signals on our chip to within 50 picoseconds, that is literally 50 trillionths of a second. And we had to go get the signals aligned to this super small granularity and do it on every chip 2 billion times a second.”

“we took over 700 FPGAs and put the entire full reticle chip on an FPGA cluster and ran a dozen different models with our full inference stack on them before the chips came back”

Patrick asks some structurally good questions - pressing on what the robotics competition analogy means for company-building, probing vertical integration limits, and surfacing competitive threats like OpenAI's Jalapeno - but he explicitly discloses being a large multi-round investor, which visibly softens the dynamic; claims go largely unchallenged and the closing act degrades into mutual admiration and a personal-touch softball closer.

“So sort of like loading the gun and then firing it. Like if I think about it in super simple terms.”

“I'm trying to ask questions that are broader and interesting and could be objections to what you're doing. And we'll keep doing that.”