Posted by todsacerdoti 1/20/2026
There's a bet here on profitability and it needs to play out.
How long do investors normally wait to see if a bet on new technology is a winner? I imagine that's quite arbitrary?
The poster child for this is Tesla. Nothing fundamental justifies Tesla's valuation.
IMHO the only rational way to look at the future of AI and the companies from profit from it is to look at geopolitics.
The market seems to have decided there's going to be one winner of the AI race. I actually don't think that'll be OpenAI. I think it'll be Google or Nvidia of the companies currently in the race. But I also don't think it'll be either of them.
The magic of software is that it is infinitely reproducible. That makes it difficult to build a wall around it. Microsoft, Facebook, Apple and Google have successfully built moats around their software very successfully in spite of this. Google's big advantage in the AI race is their ability to build and manage data centers and that they'll probably end up relying on their own hardware rather than NVidia.
I think China will be the AI winner or they'll make sure there is no winner. It's simply too important to them. For me, DeepSeek was a shot across the bow that they were going to commoditize these models.
The US blocked the export of the best lithography machines AND the best chips to China. IMHO this was a mistake. Being unable to import chips meant Chinese companies had no choice but to make their own. This created a captive market for China recreating EUV technology. Chinese companies have no choice but to buy Chinese chips.
The Chinese government has the patience and infrastructure for recreating ASML's technology and it's an issue of national security. And really all it takes is hiring a few key people to recreate that technology. So Western governments and analysts who said China will take 20+ years to catch up (if they ever do) simply don't understand China or the market they're talking about.
They sound exactly like post-WW2 generals and politicians who thought the USSR would take 20+ years to copy the atomic bomb. It took 4 years. And hydrogen bombs came even quicker.
There's a story that should get more attention: China has reportedly refused a deal for NVidia's latest chips [1]. If true, why do you think they're doing that? Because they don't want to be reliant on foreign chips. They're going to make their own.
[1]: https://ca.finance.yahoo.com/news/nvidia-stock-slides-china-...
Are they already "too big to fail"? For better or worse, they are 'all in' on AI.
I do hope they crash so that I can buy as much as possible at a discount.
In general, they often get stung by the dead cat bounce, =3
This means that cash revenue will likely remain high long after the LLM hype bubble undergoes correction. The market will eventually saturate as incremental product improvements stall, and demand rolls off rather than implodes. =3