What’s your plan for the DeepSeek induced crash in AI and semiconductor stocks?
Wait and watch is my plan. I believe eventually all the learnings from DS’ paper will help expedite the scaling of US LLMs. So the hyper scalers (msft, amzn, etc.) should be fine in the long run. However, I’m not so sure of NVDA. The need for $100Bs of Capex for GPUs may not be a valid case any longer.
Nvidia for sure is over valued for a long term investor. This hype of AI would go down eventually and with that Nvidia’s price. It is not just about if that much amount of cape is really needed or not but also about if the companies can generate enough returns with that kind of capex. What is the revenue model around it? How are these AI models planning to generate revenue? I highly doubt if many users would be willing to pay $20 per month which ChatGPT pro or even Microsoft is trying to charge. And why would someone pay that kind of money if they aren’t able to derive that much value from it. And if that is not the model then everyone would be competing for the add revenue pie which is dominated mainly by Google. One can argue that the long term bet is the gain in the efficiencies and with that the costs but that is really a long term scenario beyond ten years as it takes a very long time for widespread adoption of such tech and for the tech to mature to that level. Once the industry understands the importance of such efficiency gains then only industry would start paying for such it. So my best guess is that atleast for next 5 years this would not much reflect in the revenues and profits and once the market begins to see that in the financial results, it would begin to reflect in the share prices as well. But in the very short term let’s say for an year or so the hype would not die down so easily. One or the other positive story would keep coming out and it might be business as usual.
Selling NVDA seems the right thing to do. That’s what I’m doing. Buy the dip on MSFT, AMZN is also the best thing, as they’ll adapt fast and thrive with scaled US LLMs.
I Dug Into How DeepSeek Really Works. Have summarized it here
DeepSeek’s AI model, DeepSeek-R1, focuses on reasoning capabilities, powered by pure reinforcement learning.
Unlike traditional AI models, it doesn’t rely on large supervised datasets (that uses large computational power) but uses techniques inspired by AlphaZero, mastering tasks through self-play and optimization (same that beat every other AI in chess by Google). I wonder why Google didn’t think of this move before.
Instead of increasing model size and computation, DeepSeek emphasizes algorithmic efficiency, making high-quality AI accessible for under $6 million
DeepSeek released its model under the MIT license, fostering collaboration and further development within the AI community. Transparency in its architecture and training techniques is a bold move most companies haven’t done before.
There are concerns over data transmission to China and state-aligned outputs in sensitive areas raise (problems in China or related questions about politics or some censored events) questions about independence.
If DeepSeek’s approach becomes the industry norm, it could disrupt the AI hardware supply chain, reducing dependency on expensive GPUs. Nvidia dropped 17%, total combined drop of $1 trillion across all major stocks.
@Ansh_Orbit_448 Would recommend not to sell Nvidia right now off the bat. See what Sam Altman has to say about DeepSeek and AI in general. I agree with his perspective that world will need more AI than ever and DeepSeek’s approach may just optimize it more. It doesn’t mean they would need any less GPUs or less compute power. I believe in the long run Nvidia will still win
Agree. What may change is the quantum of GPU purchasing form the hyper scalers (msft, amzn, meta, etc.). This obviously means lower growth rates for NVDA. So, the PE will shrink for now, which I’m fine with as a long term investment.
I also have another theory. This development will also mean lots of other companies (even non-US) will buy the excess GPUs, instead of the hyper scalers. NVDA’s buyer base will be democratized.
If anything, nvidia will continue being prominent. Deepseek also runs their models on nvidia chips. And here is a video where they into detail explaining how this is just the beginning.
