From NextGen — Pick Your First Customer Carefully
At Farnborough this month, Anduril unveiled Thunder and Archer unveiled Halo, one day apart. Same airframe, same hybrid powertrain, same core systems. One is a defense aircraft. One is a commercial one. Beta did the same trick: its new MV250 is the A250 with the cockpit and cabin deleted.
That’s the pattern running throughout this issue. New hardware doesn’t get cheap by being clever. It gets cheap by finding a first buyer who’ll pay a premium and generate hard feedback while the cost curve is still ugly, then porting that learning to everyone else. Four reactors went critical by July 4. The interesting question isn’t whether they work. It’s who pays ~$214/MWh first.
If you're bringing new hardware to market, who's your first paying customer?
The Good: Manufacturing activity expanded in June for the sixth consecutive month, with the ISM Manufacturing PMI at 53.3% — the longest expansion streak since 2022. Input prices fell 9.1 points to 73.0, the largest single-month drop since July 2022.
The Mixed: PMI is down 0.7 points from May. Employment came in at 49.7%, up 1.1 points but still contracting for a 33rd straight month. Manufacturers are growing output without growing headcount.
Big Movers: IndustrialSage now tracks $1.948T in announced factory commitments across 219 companies. Eight of them are 76.4% of the total. Apple ($600B), TSMC ($265B), and Micron ($250B) alone account for $1.1T.
Our Take: The expansion is real and it is extremely narrow — semiconductors are 71% of everything tracked. Strip out chips and pharma and the manufacturing boom gets small fast. But the line to watch is employment: six straight months of expanding output against 33 straight months of contracting payrolls. Factories are being announced. Jobs aren’t showing up at the same rate. That’s either a productivity story or a warning, and which one depends entirely on whether the announced plants ever get built.
One Airframe, Two Customers
At Farnborough, Anduril revealed Thunder (defense) on July 21 and Archer revealed Halo (commercial) the next day — two variants on one clean-sheet design: series hybrid-electric powertrain, two tiltrotors, a modular payload bay. Marubeni Aerospace is launch partner for Japan. Full-scale surrogates have flown; first flight of the production aircraft is targeted for 2027.
Beta flew the first operational missions under the DOT/FAA eVTOL Integration Pilot Program on July 10, carrying United Therapeutics’ manufactured organs between Virginia and Maryland — first in a series of use cases spanning at least 26 states. At Farnborough it unveiled the MV250, its A250 with the cockpit and cabin deleted: 1,300 nm repositioning range, 2,000 lb payload.
Our Take: ‘Dual-use’ usually means a defense contractor issuing a press release about commercial applications. This is the opposite, and it’s a manufacturing decision before it’s a market one. One airframe, one powertrain, one supplier base, one line — then fork the payload bay. The defense variant gets a buyer who’ll pay for low-volume units and generate real operational feedback; the commercial variant inherits that learning without paying for it. Beta didn’t build a military aircraft, it deleted the cockpit from one it already builds. Open question: whether the variants stay close enough to keep sharing a line, or whether requirements drift until you’re funding two programs.
Four Reactors Went Critical. Now Find a Buyer.
Executive Order 14301 targeted three reactors critical by July 4, 2026. Four made it: Antares (June 4), Valar Atomics (June 18), Deployable Energy (July 1), and Aalo, whose Critical Test Reactor went critical at 12:20 a.m. on July 4 — eight months from groundbreaking to sustained chain reaction, and the first new reactor at Idaho National Laboratory in ~50 years. Radiant’s Kaleidos is fueled at INL’s DOME next, with a factory targeting ~50 reactors a year behind it. The speed has critics: DOE rewrote its safety standards and exempted the program from environmental review.
X-energy is a founding Tier 1 member of Project Prometheus, a $60M three-year initiative led by Idaho National Laboratory, NVIDIA, and AWS under the DOE’s Genesis Mission, committing its Xe-100 reactor and TRISO-X fuel designs plus $10M of its own capital. Separately, TRISO-X took an $11M Tennessee grant to expand its Oak Ridge fuel campus — sized to fuel 55 Xe-100 reactors, or ~4.5 GW.
Antares raised $470M ($370M equity, $100M debt) led by Paradigm and Caffeinated Capital. It builds TRISO-fueled reactors in the 100 kW to 1 MW range, and its Mark-0 was the first Reactor Pilot Program unit to reach criticality. Antares is one of three finalists in the Pentagon’s Advanced Nuclear Power for Installations program, testing on Air Force bases in Colorado and Montana, and targets first power next year with military deployments in 2028.
Our Take: Reaching criticality has stopped being the hard part — four teams did it in a single summer, one of them eight months from groundbreaking. Selling the electricity is the hard part. Lazard pegs new small modular reactor power at ~$214/MWh, which isn’t competitive with gas and doesn’t have to be if you pick the right first buyer. Antares picked the Pentagon, famously price-insensitive, with bases that would rather not depend on the grid. X-energy picked the AI buildout, where the binding constraint is interconnection queues, not cents per kilowatt-hour. Both are beachheads, not markets. And DOE authorization is not an NRC license.
Borrowed Channels
Zipline × Cleveland Clinic
Its first US healthcare home delivery: prescriptions first, then lab samples and inter-facility supplies, plus a consumer launch in Austin. Underneath: 2.5M+ deliveries and 135M autonomous miles without a safety incident, every flight authorized beyond visual line of sight (BVLOS).
Joby × Virgin Atlantic
A multi-year deal signed at Farnborough making Virgin its exclusive UK air taxi partner, with hubs in London and Manchester. Heathrow to central London in ~8 minutes. Joby keeps aircraft operations and CAA approvals; Virgin handles infrastructure and customer acquisition. Delta owns 49% of Virgin Atlantic.
Saronic × Samsung Heavy Industries
Autonomous vessel development, shipyard production technology, and robotics-based process automation at Saronic’s US yards, unveiled at the launch of the Korea-US Shipbuilding Partnership Center in Washington. It spans Saronic’s Franklin, Louisiana yard (another $300M going in) and its $3.2B Port Alpha facility in Brownsville, Texas.
Waymo × Uber, ending
Waymo is removing the human specialist in San Diego, Las Vegas, Tampa, and Denver, employees first; Las Vegas flipped immediately, the others need more mapping. It has also given notice it intends to exit the Uber partnership and launch through its own app in Austin and Atlanta in January 2028. Behind it: ~500,000 paid rides a week.
Our Take: None of these companies is buying capacity. They’re buying the thing that takes longest to build. Cleveland Clinic gives Zipline patient trust it couldn’t manufacture at any price. Virgin gives Joby UK customer acquisition and a Delta relationship by proxy. Samsung gives Saronic something genuinely scarce — heavy shipbuilding capability, which most US allies gave up decades ago. Then there’s Waymo, which rode Uber’s network to reach riders and is now walking away. That’s the full lifecycle in one set of updates. Borrowed channels get you to scale faster than you could build one. They also expire, usually right when you stop needing them and start competing.
Capital Watch
Sila — $300M
Led by Atreides Management and Sutter Hill Ventures, for phase two of its Moses Lake, WA anode plant. The site came online in fall 2025 at 2 GWh and is designed to reach 250 GWh over five years. Titan Silicon runs 20-40% denser than graphite; Mercedes-Benz and Panasonic are contracted, but the pitch is shifting from EVs toward AI hardware, defense, and aerospace.
Atoms - $1.7B
Led by a16z. Travis Kalanick’s holding company atop CloudKitchens plus Pronto, Anthony Levandowski’s heavy-industry automation outfit. Ben Horowitz joins the board; Uber is in the round, reconnecting Kalanick with the company that pushed him out in 2017.
Antora Energy - $550M
Series C co-led by G2 Venture Partners and Eclipse (July 30), one of the year's largest cleantech rounds; valuation undisclosed. Antora's thermal batteries store electricity as heat in solid carbon blocks — no critical minerals, around-the-clock output — aimed at industrial heat and AI data centers. New backers include Salesforce Ventures, Ribbit, and John Doerr, with Breakthrough Energy and Lowercarbon returning. Funds a second U.S. plant; a 5 GWh system for POET in South Dakota came online in under a year.Base Power - $1B
Series D at a $13B post-money valuation (announced today), a year after its ~$1B Series C. Led by Ribbit, Addition, Valor Equity Partners, and JPMorganChase's Strategic Investment Group, with Coatue, Thrive, and a16z joining. Base sells backup power and grid services on subscription rather than the backyard battery itself, and just launched Base Core, a 39.2 kWh Austin-built unit. It's installed 500+ MWh, deploys ~100 batteries a day, and runs in Texas and Illinois.Anduril - reportedly in talks at ~$100B
Roughly Northrop and Lockheed territory, two months after a $5B raise at $61B. One floated structure: investors commit to a second round at a higher valuation within a year, tied to financial benchmarks. Anduril says no decisions have been made. Revenue more than doubled to $2.2B in 2025.
Our Take: Look at what the money is chasing. Defense tech venture funding topped $12B in the first half of 2026 — more than all of 2025 — and Anduril may triple its valuation in a year on $2.2B of revenue. That’s the price-insensitive customer showing up in the capital stack, not just the order book. Sila is the tell in the other direction: it built for EVs and is now pointing the same anode at AI hardware and defense, because that’s where the buyers who don’t flinch at price are. Same plant, different customer. The risk is the obvious one — design your cost curve around a buyer who doesn’t care about cost, and you may never get down it.
The biweekly read on rebuilding American industry
Funding, factory openings, and the takes that cut through the noise, for the people building the next American manufacturing base.