Key insights
- The analyst expresses a 'no' on investing in SpaceX, Anthropic, and OpenAI at current valuations and market conditions. For SpaceX, the valuation is deemed too high relative to revenue, with a prediction of a market correction impacting all assets. For AI companies Anthropic and OpenAI, the analyst cites low defensibility due to rapid innovation, potential competition from Chinese open-weight models, and NVIDIA's advancements in local AI processing, suggesting a lack of sustainable competitive advantage.
SpaceX: Great company, terrible price. A ~$2 trillion valuation on ~$20 billion of revenue just doesn't add up for me. On top of that, the market is running very hot right now, and markets always correct — go pull up Jan–Feb 2022 and the tech crash, everything came down together. When that happens again (and it will), SpaceX comes down with everything else. I'd rather wait and pick it up closer to ~$1 trillion. This isn't "no forever," it's "no at this price."
Anthropic: The growth is genuinely insane (~$10B ARR in Jan 2026 to ~$45B ARR now), but that's exactly the problem. A year ago they were nowhere — it was all OpenAI, ChatGPT, and Gemini — and they came from behind and took the space fast. If they could do that to someone else, someone can do it to them. The threat I see most clearly is Chinese open-weight models. I think China dumps fully open models, weights and all, and people just run them locally — the way Airbnb did it: took open-source models, kept all their data in-house, nothing going to China. The "I can't run a huge model on my laptop" problem? NVIDIA's solving it — Jensen's new machine is reportedly built for agents instead of humans, with data-center-class GPU power and the speed agents actually need. Low defensibility, so I pass.
OpenAI: Same story, arguably worse. At least Anthropic has enterprise clients as some kind of moat; OpenAI really doesn't. And I'm broadly skeptical of software-only businesses in a world moving this fast.
So SpaceX, Anthropic, OpenAI — my answer on all three is no.
Tell me where I'm wrong. What's the bull case I'm missing, especially on defensibility?