This is an excellent read! And thanks for linking my article. I'll also add that Trump's tariffs (or planned tariffs) probably had quite a bit to do with the crash as well.
Thanks! Love your stuff and have been a happy subscriber for a while now. For the tariffs... it's tough to say.
Pricing is always hard to disentangle, but before the weekend, we weren't seeing significant divergences (which we now are) between companies that would be impacted or not (e.g. NVIDIA/GM vs Meta/Alphabet). I'd largely argue that markets (in aggregate) didn't really "believe" in the tariffs, whereas this Monday... much more so (even with Mexico being "delayed" now by a month). If you were betting on tariffs, you'd probably be executing this trade—and in aggregate, markets weren't.
To be fair to markets though, this administration could very well (even now) announce some kind of deal and say, "Ok, we win and tariffs are successful and no longer going into effect!"... or double them to 50% in response to retaliatory tariffs from Canada/Mexico. Given that, you would trust his serious-minded treasury secretary much more that it's mostly bluster (at least before the weekend and Trump's clarifications to make it clear he's "doing it").
Partly why I wanted to get this particular piece out on Sunday—I was pretty sure that by market open today, whatever I said would be swamped by tariff news!
Break many AI bubbles and expand the AI market. Especially, OpenAI is already a lagging company. In 2025, OpenAI's market share will drop significantly.
The bizarre thing about the NVDA story was the delay between the recent Deepseek release, around January 14, and the sudden market reaction to it on January 27.
A further puzzling question is why the markets are apparently ignoring the other LLMs from China including Qwen (https://chat.qwenlm.ai/) and Yi (https://huggingface.co/01-ai/Yi-1.5-34B-Chat). Both claim to be comparable or better than ChatGPT and claim to have been trained for small fractions of the cost.
It seems that US markets are still irrationally resistant to the notion that China does AI at least as well and at far lower cost than the US. It's difficult to believe that, as has been suggested, these companies are blatantly lying about their costs or that they somehow sent billions of requests to ChatGPT to train their systems without OpenAI's knowledge.
China has demonstrated that a group of smart engineers, working anywhere in the world, can find efficiencies and improve algorithms and lower costs. As with all technologies, that process will inevitably continue. In two years the cost to train and run an LLM will be far lower than today.
But even after its recent decline, NVDA still has a Price/Earnings ratio of 47 which implies that it will continue at its current growth rate for many years. How absurd!
The market is running on hype and ignoring the facts. Dramatically falling requirements for their chips, and the lack of any real-world applications of LLMs, would normally cause NVDA to crash. But at the moment the disdain for China and the unwillingness to recognize China's technological prowess are keeping NVDA at its current level.
When the market finally realizes all this, NVDA will close below $50 by the end of 2025.
This is an excellent read! And thanks for linking my article. I'll also add that Trump's tariffs (or planned tariffs) probably had quite a bit to do with the crash as well.
Thanks! Love your stuff and have been a happy subscriber for a while now. For the tariffs... it's tough to say.
Pricing is always hard to disentangle, but before the weekend, we weren't seeing significant divergences (which we now are) between companies that would be impacted or not (e.g. NVIDIA/GM vs Meta/Alphabet). I'd largely argue that markets (in aggregate) didn't really "believe" in the tariffs, whereas this Monday... much more so (even with Mexico being "delayed" now by a month). If you were betting on tariffs, you'd probably be executing this trade—and in aggregate, markets weren't.
To be fair to markets though, this administration could very well (even now) announce some kind of deal and say, "Ok, we win and tariffs are successful and no longer going into effect!"... or double them to 50% in response to retaliatory tariffs from Canada/Mexico. Given that, you would trust his serious-minded treasury secretary much more that it's mostly bluster (at least before the weekend and Trump's clarifications to make it clear he's "doing it").
Partly why I wanted to get this particular piece out on Sunday—I was pretty sure that by market open today, whatever I said would be swamped by tariff news!
Clearly argued, cutting through current jargon around the issue James, thank you.
Break many AI bubbles and expand the AI market. Especially, OpenAI is already a lagging company. In 2025, OpenAI's market share will drop significantly.
https://aidisruption.ai/p/deepseek-confirms-my-doubts-about?r=2ajqea
The bizarre thing about the NVDA story was the delay between the recent Deepseek release, around January 14, and the sudden market reaction to it on January 27.
A further puzzling question is why the markets are apparently ignoring the other LLMs from China including Qwen (https://chat.qwenlm.ai/) and Yi (https://huggingface.co/01-ai/Yi-1.5-34B-Chat). Both claim to be comparable or better than ChatGPT and claim to have been trained for small fractions of the cost.
It seems that US markets are still irrationally resistant to the notion that China does AI at least as well and at far lower cost than the US. It's difficult to believe that, as has been suggested, these companies are blatantly lying about their costs or that they somehow sent billions of requests to ChatGPT to train their systems without OpenAI's knowledge.
China has demonstrated that a group of smart engineers, working anywhere in the world, can find efficiencies and improve algorithms and lower costs. As with all technologies, that process will inevitably continue. In two years the cost to train and run an LLM will be far lower than today.
But even after its recent decline, NVDA still has a Price/Earnings ratio of 47 which implies that it will continue at its current growth rate for many years. How absurd!
The market is running on hype and ignoring the facts. Dramatically falling requirements for their chips, and the lack of any real-world applications of LLMs, would normally cause NVDA to crash. But at the moment the disdain for China and the unwillingness to recognize China's technological prowess are keeping NVDA at its current level.
When the market finally realizes all this, NVDA will close below $50 by the end of 2025.