A Rising Tide Won't Raise a Sinking Ship
Another one on AI.
Random thoughts:
- Claude Mythos was a wonderful 48 hours, I feel like I've been lobotomised since it was banned.
- Every portfolio is an unintended AI bet.
- Credit markets are going to get tonnes of AI capex driven supply.
- Infrastructure is one big ol' AI bet.
- What's been driving equities? A chunk of it is the 3% of MSCI World that is represented by memory stocks. I can see this sector becoming almost 10% of MSCI world in the next 2 years.
- Trying to find convexity remains challenging...
- With the Mythos ban, does this mean the US Government is going to regulate knowledge to a certain level?
- For the consumer, does this mean you cannot access more than a model that is more than 1.5x better than today?
- Any ban or limiter is utterly devastating for the AI complex, allows other players (ahem, China) to catch up rapidly.
- Open source is going to screw closed source models - therefore the hardware layer will win until the software/AI winners become apparent.
- Intel is the most important company in the US. It allows the US to remove the reliance on Taiwan and reduce geopolitical risk completely.
- It is not priced as the most important company in the US.
- SpaceX making its own semi-conductors is the only other business to challenge this notion.
My laptop today can run a better model than you could access via ChatGPT/Claude/Gemini 12 months ago.
The rate of change remains exponential which is scary, and I am comfortable saying every frontier model is more than twice as good than a year ago.
From an investment perspective, everything is being underwritten by the hyperscalers/Mag-7 who are experiencing incremental returns. Anything AI touches is seemingly an NPV+ projects and we're finally getting to the point where the Mag-7 are going to tap credit markets.
This means your credit exposure will also become incrementally a larger AI bet as it is one of few areas where credit market issuances will rise.
Uncertainty
Questions to answer from here.
- How will credits evolve? Lots of supply to come into markets now, so where will corporate credit trade relative to government bonds. Both have decent of supply coming on in the near term.
- So many investors remain underweight AI/Tech - if investors don't capitulate to allocate here, does that mean markets keep ripping?
- OpenAI & Anthropic IPOs are going to be interesting, will the Mag-8 (SpaceX now included) then become the Mag 10?
Not enough people paying attention to potential US AI regulation, the whole economy is dependent on the AI trade. What is more important, capitalism or stability?
Everything has a cost, so will the US really give up the AI lead to make AI "safer" by restricting access to the best models? Or do they knight a small group of companies as their AI champions (OpenAI, Google, SpaceX) to whom they will delegate "safety", while those that are not knighted (Anthropic) lose?
The AI trade is split in 2, the OpenAI ecosystem and the Anthropic ecosystem. There are certain listed companies that are directly linked to these ecosystems, the microstructure performance in markets can be directly linked to which frontier model is doing better at any given point in time.
For one ecosystem to be knighted over another is going to create some volatilty. Sam (OpenAI, >$US1T) and Demis (Google, >$US4.5T) sat next to Trump at the G7 lunch while Dario (Anthropic, >$US1.5T) was forced to sit across the room away from the US delegation.
There are going to be winners and losers, sooner rather than later.
A rising tide won't be able to raise a sinking ship.