The Physical Layer

The ranking

Ranked by the most recently disclosed private valuation, the cohort is led by a company that did not exist five years ago. Saronic, which builds autonomous surface vessels for the US Navy, closed a $1.75bn Series D at $9.25bn at the end of March — more than double the ≈$4bn it carried a year earlier. Physical Intelligence, a robot-foundation-model lab, follows at $5.6bn; Lightmatter, in silicon photonics, at $4.4bn. Below them the marks thin quickly. Physical Intelligence’s figure may already be stale: Bloomberg reported in March that the company was in talks to raise about $1bn at more than $11bn, and no close had been announced as at the date of this note. We rank on what has printed, not on what is being discussed.

Three businesses — Isomorphic Labs, Synchron and Panthalassa — have raised substantial capital without a cleanly disclosed post-money. The distinction matters more than it sounds. A priced round fixes a number that every subsequent holder, auditor and buyer can reference; a large cheque from a well-known name signals conviction but leaves the number itself unfixed. We treat the two differently throughout.

Capital into the frontier

The flow of primary capital is heavier than the marks alone suggest. Isomorphic Labs, the Alphabet-backed drug-design business spun out of DeepMind, raised $2.1bn in May without naming a valuation — the single largest cheque in the cohort. Saronic’s $1.75bn is the second. After those two, round sizes fall to the low hundreds of millions. The concentration of capital mirrors the concentration of marks: a short head and a long tail. Two companies took three-quarters of the money.

Grouped by domain, life sciences and defence absorb roughly three-quarters of the eighteen-month total. Robotics, energy, neurotechnology and ocean compute share the remainder. The breadth of the “AI meets atoms” story is real, but the money is not evenly spread.

The scale

Plotted against the capital each has consumed, the marks show what is actually being underwritten. There is no revenue line to anchor these valuations, because the businesses are pre-commercial. What the chart shows instead is a money-in multiple — the mark divided by every dollar the company has ever raised. It is a crude measure, but it is the only one available when there are no sales to divide by, and it says something useful: how much value the market is assigning above the cost of the assets and the team assembled so far.

The multiples cluster tightly. Synchron sits at roughly 2.9× money-in, Precision Neuroscience 3.2×, Radiant 3.4× and Saronic 3.6×; Lightmatter and Physical Intelligence both around 5.2×, and Panthalassa 5.4×. A cohort spanning warships, reactors, surgical implants and photonic chips is being priced within a band of roughly three to five and a half times capital consumed. That consistency is itself the finding: investors are not underwriting seven different businesses so much as one recurring bet.

What that bet looks like when it matures is visible one tier up. Anduril, the same thesis a decade older, was marked at $61bn in May on roughly $2.2bn of 2025 revenue — about 28× sales, against the 1.6× to 2.7× at which the listed defence primes trade. The frontier is being priced today on the business that Anduril became. Whether the next tier gets there is the open question.

What the desk is watching

Three observations:

1. None of these businesses is yet valued on revenue, so each mark is a claim about execution that has not happened yet.
   • Isomorphic holds partnerships with Eli Lilly, Novartis and Johnson & Johnson worth close to $3bn in potential milestones, yet no patient had been dosed with a drug it designed as at May 2026, and first-in-human trials have already slipped from end-2025 to end-2026.
   • Saronic says it will be able to build around twenty of its 180-foot Marauder vessels in 2027 at full capacity, and in July selected Brownsville, Texas for Port Alpha, a $3bn shipyard due to open in 2028.
   • Radiant took delivery of its first tranche of TRISO fuel at Idaho National Laboratory on 1 July, but its Kaleidos reactor did not reach criticality by the federal 4 July target and its five-phase test campaign now runs through the summer.
   • Precision Neuroscience is the one name with a cleared, saleable product, and it carries the smallest mark in the cohort. Delivery and price are not yet moving together.
2. Atoms are heavier than software. The gap between a financing and a shipped product is longer, and the capital required to cross it is larger. A cohort raising at this pace is a cohort that will need to keep raising, which makes each of these marks a bet on the availability of the next round as much as on the technology.
3. Isomorphic, Synchron and Panthalassa were each funded without a clean post-money. A large cheque from a marquee backer is a strong signal, but it is not a priced valuation, and we flag the distinction wherever it applies.