I’m generally skeptical of defense-tech valuations that double in a matter of months, because most of them are pricing in a procurement win that hasn’t happened yet rather than anything that shipped. Anduril going from a $61 billion round in May to talks around $100 billion now reads, on the surface, like exactly that pattern.
I don’t think it’s just hype this time, and the reason is a report that has nothing to do with Anduril at all.
The UK AI Security Institute and the US CAISI ran a joint preliminary evaluation and found Kimi K3, the leading Chinese open-weight model, trailing US frontier closed models on cyber capability. That’s the first quasi-official, cross-government measurement of the gap Anduril’s entire valuation thesis depends on, and almost nobody has connected the two stories.
What Is Anduril’s Round Actually Pricing?
Ask what a defense-tech investor is buying at $100 billion that they weren’t buying at $61 billion three months ago. Not new contracts announced in that window — the jump is too fast for procurement cycles to explain it. What’s changed is the market’s confidence that Western AI-enabled defense systems hold a durable capability edge over what a peer competitor can field, and that edge is precisely what the AISI/CAISI evaluation just tried to put a number on. Anduril’s valuation is a bet on a gap; the joint evaluation is the closest thing to a referee that gap has ever had.
A Government Evaluation Investors Aren’t Using
Venture rounds in defense tech are usually priced off classified assessments nobody outside the deal can see, which is part of why the valuations feel untethered. The AISI/CAISI finding is unusual precisely because it’s public, specific to cyber capability, and jointly signed by two allied government bodies with no commercial stake in Anduril’s round. Treating it as irrelevant to a $100 billion price tag because it evaluated a chatbot rather than a drone swarm misses the point: it’s the first real data point on whether the underlying AI capability gap Anduril is priced on actually exists at the scale the round assumes.
Where the Evaluation Falls Short of the Valuation
Cyber capability in a general-purpose model is not the same thing as autonomous weapons capability in a purpose-built defense system, and it would be a mistake to read Kimi K3 trailing on one narrow benchmark as proof that the broader gap Anduril is priced on is as wide as the round implies. The evaluation is preliminary, narrow, and silent on everything that actually determines a battlefield outcome.
Still, a $39 billion jump in three months deserves a better citation than “trust us.” Now, for the first time, there’s one on the record.
Leave a Reply