Summary:
I bought into Nvidia ($NVDA) at around $118, right after the DeepSeek announcement. The market tanked on this announcement and Nvidia dropped nearly 17% in a single day, along with the rest of the AI sector. Yet I found this to be a complete overreaction because Nvidia chips were still powering DeepSeek, even though the media spun it as a threat.
DeepSeek is a Chinese AI startup that launched its R1 model on January 27, 2025. It quickly reached #1 on the U.S. App Store without heavy infrastructure investment. They reportedly used Nvidia A100/H800 chips, older-generation hardware allowed to export to China by U.S. export controls.
My Thesis
- They used Nvidia chips even if they were older ones. If DeepSeek was threatening Nvidia, it wouldn't be deploying Nvidia hardware. This reinforced the idea that Nvidia's infrastructure is still core to global AI. Much of Nvidia’s value doesn’t just lie in its chip performance but also it’s software ecosystem (CUDA, cuDNN, etc.). Their software ecosystem remains indispensable for training and inference. DeepSeek may innovate at the model layer, but they still rely heavily on Nvidia’s stack to deploy at scale.
- DeepSeek is a “test-time scaling” showcase, not a threat to Nvidia supply chains. Nvidia itself called DeepSeek “an excellent AI advancement” and dismissed existential threats.
- Jeavons Paradox: cheaper, more efficient AI means more AI demand. Industry analysts believed DeepSeek would expand the AI pie, not shrink it.
- The fundamentals didn’t change. Nvidia still remained the dominant AI-infrastructure provider and major AI platforms like Google and Microsoft continued to invest in Nvidia chips post-dip.
Price Target
My target is the pre-crash highs at around $142 per share. I don’t expect to DCA further. I expect a sharp rebound tomorrow (opportunity to buy calls) and then to consolidate for a week or so, assuming no more unexpected news comes out. Buy fear, ride fundamentals.