Defence got the round
A 140% valuation increase in twelve months is not a normal venture outcome. It is what happens when the customer is a government and the switching cost is close to absolute.
Why defence multiples look wrong
They look wrong against commercial software comparables because the underlying business has different properties. The buyer is not price-sensitive in the way an enterprise procurement office is. The integration, once made, is close to permanent. And a single programme award re-rates the entire book in a way that recurring SaaS revenue never does.
Those properties support multiples that would be indefensible in a competitive commercial market, and they are real properties rather than narrative ones.
The risk that does not show up in the multiple
Revenue concentrated in government programmes is revenue exposed to programme cancellation, budget cycles, and political change. None of those correlate with product quality, which means the usual venture logic — build a better thing, capture more market — is only partially load-bearing.
The switching costs that make the multiple defensible are the same switching costs that make a cancellation unrecoverable.
Where this sits in the flow
MGX has closed a $49bn AI fund, in a half-year that set a record at $510bn of global startup investment. Vehicles that size cannot deploy into seed without owning the stage, so the capital goes late and into infrastructure.
Defence autonomy is one of a small number of segments absorbing that capital at these prices. The others are compute, healthcare, legal tech, and regulated business tooling — all of them categories where the buyer is constrained by something other than preference.
The gap nobody reconciles
The same coverage that reports $49bn funds also reports that most founders still need proof rather than narrative to raise. Those facts are not in tension. Enormous capital concentrating at the top is entirely compatible with a tightening at the bottom, and historically accompanies it.
The structural question is what a fund that size does if frontier-lab valuations stop compounding. There is not enough late-stage AI supply to absorb capital at this rate on current terms. Either the terms move, or the definition of an AI investment gets a great deal broader.
Crescendo AI — Latest AI Startup Funding News and VC Investment Deals — 2026 → · Crunchbase News — Global Startup Investment Hit Record $510B In H1 2026 → · Crunchbase News — North American Startup Funding Shattered Records In First Half Of 2026 →