Hadrian has raised about $2 billion since it started selling automated factories to the defence industry. On August 6 it closed the biggest cheque yet: a $1.37 billion Series D at a valuation just shy of $8 billion, about five times its valuation a year earlier. The company builds the manufacturing infrastructure that weapons programmes stand on, not the weapons themselves, and that distinction is what investors paid for.
The factory builder pairs automated plants with proprietary software called Opus and counts Lockheed Martin, RTX and Anduril among its customers
In March 2026 it opened a fourth site in Alabama to mass-produce submarine components under a $2.4B public-private partnership
This is a wager that automated production capacity, not any single component, is the durable strategic asset
WCM Investment Management, Washington Harbour Partners, Valor Equity Partners, 137 Ventures and Baillie Gifford joined JPMorgan in leading the round. Participating investors included Founders Fund, Lux Capital, Andreessen Horowitz, CapitalG, Altimeter and T. Rowe Price. The Series D landed a little more than a year after Series C, and the gap between the two rounds is the real data point: the company's valuation roughly quintupled as its contract backlog was formalised into state partnerships.
A factory company, not a weapons company
Most of the money in defence technology over the past three years went to companies designing things: uncrewed aircraft, drone swarms, battlefield AI. Hadrian does none of that. It builds the production infrastructure that the designers depend on.
Its model is factories-as-a-service: precision machining, robotic cells, computer vision and the Opus platform layer that routes jobs across the floor. Founder Chris Power calls the plants "highly automated, design-agnostic factories." The same cells can turn out a satellite bracket, a submarine fitting or an aircraft component without retooling for each one.
Largest raise in US defence manufacturing
Just shy of $8B, post-money, on the Series D close · TechCrunch, 2026
Three million square feet, four sites
What the capital buys is physical: just under three million square feet of production capacity across four sites. The newest, in Alabama, opened in March to mass-produce submarine components for the Virginia-class and Columbia-class programmes. The United States Navy designated the site as its high-volume submarine line, and the deal was structured as a public-private partnership valued at $2.4 billion.
Hadrian's customers sit at the top of the defence-design chain: Lockheed Martin, RTX and newer entrants like Anduril. The pattern of the last three years was venture money into design. Hadrian inverted it: the capital now buys the factory floor itself.
Four automated sites running Opus
Precision machining plus software across the whole footprint · Dealroom, 2026
Why a private factory is now a strategic asset
This reads better as an asset-class shift than a funding story. Three years of defence venture money funded autonomy and software; the next three are funding the factories those systems need. Output capacity is hard to reproduce, slow to scale, and tied to long procurement cycles. Those are the same properties that make hard assets attractive during a production squeeze.
Being able to produce — and to have advanced factories — is deterrence.Chris Power, founder and CEO, Hadrian
That sentence is the investment thesis in a frame. When a state signs a $2.4 billion partnership for a factory floor, capacity becomes a national production line, and the price of the floor behaves like a hard asset again.
Reading the valuation right
| Metric | Series C (2025) | Series D (2026) |
|---|---|---|
| Amount | $260M | $1.37B |
| Valuation | ~$1.5B | $7.87B |
| Lead | Founders Fund, Lux | JPMorgan SGI + 5 co-investors |
| Signal | US growth scale | submarine partnership |
The real question is what the next round pays for. Of the $1.37 billion, the larger part funds capacity expansion and the software layer that measures its output. Which part carries the actual value?
What does the factory model do to the price of the software layer?
Probability: 70% over the next ten years.
Data on machine state, failure modes and cycle-time totals accrues with the software provider. Retained operating data is a liability when stored badly and an asset when it drives reliability contracts.
✅ Arguments for
a $2.4B submarine partnership gives the software layer a recurring-data jurisdiction;
and defence backlogs stretch far enough to underwrite the next factory round.
Confirmation criteria: follow-on submarine and aviation production, with the Opus layer booking licence revenue beyond the factory walls.
❌ Arguments against
the automation premium only pays if utilisation stays high across all four sites;
and a moat built on software is weakened the moment the top-primes build the same stack in-house.
Disconfirmation criteria: no new production contracts turn up within two quarters, or a prime swallows the factory+software pair itself.
Key signals to track
US Navy contract schedule and follow-on submarine tenders
Whether the Opus layer licenses outside its own factory walls
Utilisation across the Alabama and third and fourth sites
How the post-money terms convert into backlog and production data
Development scenarios
The system is converging on a model where the factory is the bottleneck, not the design.
🟢 Development scenario (25%)
Implications: the production layer re-priced up toward the hardware-as-a-service label, and the $2.4B partnership book pays the design house model forward.
🟡 Base case (60%)
Implications: margins hold steady, the platform softens, and the brand consolidates as the production standard.
🔴 Down case (15%)
Implications: the $7.87B mark stubbornly re-flatlines toward a classic metal-bender multiple, and the design-chain question stands apart.
As we wrote in July
We wrote in July that the defence-tech centre of gravity is shifting from design to capacity. First Anduril's industrial philosophy, now Hadrian's floor. That reading still holds, with one twist: Hadrian built its own boardroom on the software layer, not just the robots.