

This is pure speculation but I’m 90% sure that the NVidia drop had nothing to do with Deepseek R1 and was actually insider trading on the “news” that Trump was considering a tariff on Taiwanese chips.
First, R1 was released on the 20th, 5 days before the stock dropped. It wasn’t at all a secret, basically it was the talk of the town that whole week and their capabilities claims were shown to be solid very early on by many many people running independent benchmarks. But the market didn’t react.
Second, the big AI companies want all the compute they can get, they aren’t satisfied with training 10 or 100 or 1000 times more quickly, this is why they’re talking about trillion dollar data centers with nuclear reactors. Also of note, R1 was trained on Nvidia TPUs with the same amount of vram as the H100s. You couldn’t cheaply train such a model on any other brand of hardware, demand for Nvidia products isn’t going anywhere.
Third, if anything it’s the AI software companies that would take a big drop, they’re the ones who are supposedly spooked and scrambling to replicate R1 internally. The major software only players took only a small hit but recovered quickly, that would be Microsoft and Meta. Google is also a hardware company, they’re trying to move some of their chip fabs to TMSC but their TPUs are made by Samsung. They took a small hit and have not yet recovered. AMD is a hardware company, they have fabs all over including sourcing from TMSC, same story. Intel, a similar company, no change whatsoever, they don’t use TMSC at all. Nvidia took the big one, and they get ALL of their chips from… TMSC. All the action happened about simultaneously in after hours weekend trading.
Fourth, when the tariff news dropped the market seemed to be unaffected almost as if it had already been priced in over weekend trading.
Alright, changing my “taller than the average man in Napoleonic times” talking point to this one which seems way more relevant