AI AI AI
Everyone is talking about AI!
I have had a few clients interested in learning my thoughts on AI after the new developments from DeepSeek out of China.
I will be putting together more of my thoughts in my next letter.
My high-level view currently is that most AI productivity improvements will accrue to society at large, based on the recent AI open-source developments. Good news, I presume, for those who believe in the view of technology-led deflation (similar to the Internet, smartphones, etc.).
In terms of beneficiaries, my current view is that TSMC, ASML, and Nvidia are the only real, knowable long-term winners.
The rest of the FAANGs may end up spending their $300-$500 billion of capital expenditures in a race to the bottom. There will likely be winners in the software space, but we just have to watch it all very closely.
SoftBank’s investment of $40 billion into OpenAI at a $340 billion valuation will likely be a short-term top signal. I can’t understand how any fiduciary (Thrive Capital, Masa, Josh Kushner) can deploy such a massive amount with so much uncertainty about open source. Perhaps they are seeing something via their diligence that I am not seeing.
DeepSeek is open source, and I am reading that it is even more transparent than the open-source Facebook-backed LLaMA models. Ironically, a great gift to society from China!
There remain many unanswered questions regarding open source, licensing, ownership of data, API data, public data on the Internet, synthetic data, and who has the rights. It is likely true that DeepSeek used OpenAI API data, but I am not sure what OpenAI can do about it.
My sense of AI is that it is overhyped in terms of what happens in the next 1-3 years but underhyped in terms of what happens in years 10-30.
Remember, Amazon stock was almost flat from 2000-2010 before going up 40x+ from 2010-2025!
AI, like nuclear bombs, probably does represent the endgame for humanity as we know it, but I don’t know if this endgame happens in 2030, 2130, 2500, or beyond!
Since this requires a much longer conversation and a more detailed letter, I have decided to point you to some relatively good and thoughtful videos on AI.
The first interview is with Lex Fridman and Dylan Patel, followed by two videos with Dylan Patel and one with Gavin Baker at Atreides and Antonio Gracias at Valor.
Dylan is the authority on this space, in my opinion.
Please see below:
DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast
Dylan Patel is the founder of SemiAnalysis, a research & analysis company specializing in semiconductors, GPUs, CPUs, and AI hardware. Nathan Lambert is a research scientist at the Allen Institute for AI (AI2) and the author of a blog on AI called Interconnects.
Dylan and his team at SemiAnalysis (https://semianalysis.com) put out some of the best technology research I have ever seen on Wall Street. Kudos to Dylan for his unrelenting focus on AI and semiconductors and for sharing his thoughts with the world.
This video interview is 5 hours long!