DeepSeek, ChatGPT, ASML Earnings & Options
The recent Deepseek release seems to have had an impact on ASML, Chatgpt and the entire semiconductor space.
The reasoning goes something like this:
Deepseek is 10x cheaper than Chatgpt and performs comparably on benchmarks
Chatgpt is too “expensive”.
Chips which have been the “picks and shovels” of the AI gold rush are now overvalued since there are now vastly cheaper options given much smaller computer power capacities.
Consider me bamboozled because to me - this could not be wrong.
Putting asides the thought process on whether DeepSeek or Chatgpt is better -when in history has humanity ever had a deeply desired, advanced capability, and then wanted less of it when it became cheaper?
Name me an instance without being a sarcastic, self-serving dick, and I can tell you there’s none.
When food became more abundant - people ate more, not less.
When oil became more easily available - there was an arms race to get more. When energy became cheaper, people used more of it, not less.
When computational power experienced exponential increases - people used more of it, not less.
When air travel was first invented, it was a novelty. Then it became mainstream because everyone wanted one and then it became a shitty industry to invest in with next to no margins for most players because there was so damn much of it. People wanted more.
If it was useful, desirable or did something better - we always wanted more. Deepseek might kill Chatgpt - that much is true. But whichever AI service comes out on top, chips will still keep slamming in.
You discovered a software architecture that makes it cheaper to operate - so what?
Are you telling me that simply because my oil is now 10x more efficient that I would use the engine of the car any less? Hell no.
Circumstances permitting - I’m going from driving a little to driving a lot if I enjoy driving.
In the case of AI capabilities, demand will always outstrip supply. There are an infinite number of things we want AI to do for us and not one single business entity or person has ever gone, “oh yeah, we need less AI usage for the various things AI is good at.” , or, “Oh yeah, I wish AI was more expensive to deploy. It’s so cheap I wish I could use less of it.”
The cost of a useful process getting cheaper has never been an indicator of it getting used less, but rather used more.
If airplane travel was 10x less expensive today, you can bet your as there would be far more travellers come the relevant seasons. We already observe this phenomenon!
IMO - the situation boils down into the various outcomes:
Deepseek is better. Chatgpt dies or is left behind / becomes irrelevant. AI remains a goldrush with capabilities advancing more quickly.
Deepseek is not better. They fight it out. Win-win for consumers as capabilites develop.
Chatgpt is better, Deepseek dies or is left behind / becomes irrelevant. AI remains a goldrush.
None of the 3 scenarios entails chip demand being curtailed. All of it entails ongoing usage. Like the humorous Parkinson’s Law. AI demand will almost always be pegged to available supply and up.
Only a few scenarios will end the semiconductor drive:
The capital cycle of investing in semiconductors dries up due to capex restrictions
The available applications of AI grant less and less unit economic advantages in time spent, resources gained from the use of it
Quantum computing becomes commercially available for the public users, at which point, human society changes as a whole
Points #1 and #2 might be closer than #3 and far more likely but #3 ain’t impossible to rule out.
I went into this week short the $745 call (covered) for $24.10. My effective average was $717.50. The stock’s now at $683.50.
The best possible outcome is that earnings is good and share prices push back to 700 range. If it’s not great, prices back down a little.
A colosally bad earnings release would be disastrous and maybe see the stock dip to $500+.
Does it matter in the long run? Probably not. I’d lose the ability to generate “income” with this particular stock but there are plenty of other ideas to monetise.
ASML 0.00%↑ ‘s niche as the only EUV producer in the world (and no, it’s not that simple to make them, go read) is unlikely to be threatened by anything but quantum computing over the next decade.
I can’t wait for volatility to remain elevated and to see how all this plays out over the next week.
To those whom it matters - Happy Chinese New Year! Gong Xi Fa Cai!
Regards,
Irving