Trading firms design custom inference chips
Quantitative trading firms are increasingly designing proprietary in-house inference chips to optimize latency and bypass NVIDIA's premium pricing.
The argument
The guest noted that designing inference chips is a simpler technological challenge than training chips, leading many peer trading firms to build hardware teams and partner with custom silicon designers like Broadcom.
The thesis, stress-tested
✓ What validates it
- ✓Broadcom reporting increased custom ASIC revenues from financial sector clients
- ✓Trading firms publicly announcing proprietary chip deployments
▸ Risks discussed
- ▸High upfront R&D costs for custom silicon development
- ▸NVIDIA's rapid product cycle outcompeting in-house designs
Hear it yourself
"I'm not sure if such an offering exists at any reasonable price. Like, if it from maybe someone will give up a lease and I could snag it. But I think if I went to the market and tried to get a coin Wait. Sorry. Just to be clear, the chips are available, but not the capacity. I think if I had power, I could get the chips, Blackwell chips for delivery this year, but I do not think I could get the whole solution. And then if you go into 2027 for the next generation of GPUs, the Rubin GPUs, they, at least for the first, like, stretch, are gonna be very much sold out."
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE