Analysis
Add up August's largest named venture rounds, as tracked across Pulse's own coverage this month, and a pattern falls out that's easy to miss story-by-story: physical infrastructure is out-raising the application layer, by a wide margin. Four rounds account for the bulk of it:
- Castelion -- $1B Series C at a $13B valuation, hypersonic missile manufacturing
- Etched -- $700M round, valuation doubled to $21B, AI inference chips
- Base Power -- $1B Series D at a $13B valuation, US-made home batteries for grid capacity
- Groq -- $350M, pivoting from chips to neocloud infrastructure
“The read for allocators isn't that software is dying -- application-layer rounds are simply smaller and more numerous, the way they've always been.”
Together that's more than $3.2 billion in four rounds, none of which ship a product a user opens in a browser.
Compare that to the application layer's biggest August names covered on Pulse: Owner's raise for restaurant operations software and Emerald AI's round balancing data-center power loads -- itself arguably infrastructure-adjacent -- total roughly $390 million between them, a fraction of the physical-layer sum.
The read for allocators isn't that software is dying -- application-layer rounds are simply smaller and more numerous, the way they've always been. It's that the biggest single checks in 2026 keep going to things that take years to build, can't be re-prompted into existence overnight, and compound as a moat precisely because a well-funded competitor can't just ship a better version next quarter. That's a different risk-and-duration profile than a typical SaaS Series A, and LPs underwriting funds that lean into physical AI should expect longer holding periods before these bets show liquidity, not faster ones.
