It's a reflection of expectations about the future economy. Obviously, such expectations are not always accurate because humans are quite fallible when trying to predict the future. This is even more true when there is a lot of hype about a certain product.
Yesterdays price of (say) NVidia was based on the expectation that companies would need to buy N billion of USD of GPUs per year. Now Deepseek comes out and makes a point that N/10 would be enough. From there it can go two ways:
- NVidia's expected future sales drop by 90%.
- The reduced price for LLMs should allow companies to push AI into markets that were previously not cost effective. Maybe this can 10x the total available market, but since the estimated total available market was already ~everything (due to hype) that seems unlikely.
- NVidia finds another usecase for GPUs to offset the reduced demand from AI companies.
In practice, it will probably be some combination of all three. The real problems are not caused for the "shovel sellers" but for companies like OpenAI and Anthropic, who now suddenly have to compete against a competitor that can produce the same product at (apparently) a fraction of the price.
> OpenAI and Anthropic, who now suddenly have to compete against a competitor that can produce the same product at (apparently) a fraction of the price.
OpenAI and Anthropic can react by adopting DeepSeek's compute enhancements and using them to build even better models. AI training is still very clearly compute-limited from their POV (they have more data than they know what to do with already, and training "reasoning"/chains-of-thought requires a lot of reinforcement learning which is especially hard) so any improvement in compute efficiency is great news no matter where it comes from.
I think it is more expectation about expectation. You buy/sell based on whether you expect other people to expect earn or lose. It is self-referential, hence irrational. If a new play enters and peoples expectations shift, that affects your expectation of value even though the companies involved are not immediately or directly affects.
As already mentioned elsewhere, Jevon's Paradox will increase demand subsequent to improved efficiency. Yes, will not can.
So if the stock market was reflective of the economy (future or the present) then stocks should go up, instead they're going down. Why? Because the stock market is not reflective of the economy.
The stock market is essentially a reflection of societal perception. DJT which was brought up earlier is a great example, because the price of DJT has next to nothing to do with Trump's businesses and almost everything to do with how he is perceived (and remember there is no such thing as bad publicity).
Personally I think the fall will be momentary and followed shortly by a climb to recovery and beyond, but who really knows.
If you don't want to lose your money: Don't let the sensationalist financial journalists and pundits get to you, don't let big red numbers in your portfolio scare you, ignore traders (they all lose their money), don't sell your stocks unless you actually need that money for something right now, re-read your investment manifesto if you have one, and maybe buy the dip for shits and giggles if you have some spare cash laying around.
I agree that it will improve demand for AI services. There's no hard rule that the demand increase will be larger than the efficiency increase though, and so total sales of GPUs may still decrease as a result.
Yesterdays price of (say) NVidia was based on the expectation that companies would need to buy N billion of USD of GPUs per year. Now Deepseek comes out and makes a point that N/10 would be enough. From there it can go two ways:
- NVidia's expected future sales drop by 90%.
- The reduced price for LLMs should allow companies to push AI into markets that were previously not cost effective. Maybe this can 10x the total available market, but since the estimated total available market was already ~everything (due to hype) that seems unlikely.
- NVidia finds another usecase for GPUs to offset the reduced demand from AI companies.
In practice, it will probably be some combination of all three. The real problems are not caused for the "shovel sellers" but for companies like OpenAI and Anthropic, who now suddenly have to compete against a competitor that can produce the same product at (apparently) a fraction of the price.