Yesterday's News

A short essay on something that happened the day before.

Fabless and Lethal

Two Silicon Valley executives have used AI to build a billion-dollar weapons company. The AI is new. The business model and the strategic thinking behind it are not.

Yesterday, WIRED reported that Tiberius Aerospace had raised $85 million, giving the weapons start-up a valuation of more than $1 billion.1 Founded by two technology executives with no background in the arms industry, Tiberius makes AI-optimised artillery for Ukraine. One of its founders, Chad Steelberg, was previously CEO of Veritone, an enterprise AI company. To help Ukraine, he built an AI model and asked it what to build.

»At Tiberius, the questions of what the war needs and what the market will buy are answered inside the same company, by the same model«

The model’s answer was specific. Better defensive technology, it concluded, would only harden the stalemate. What Ukraine needed was greater offensive capability: artillery with enough range to keep Ukrainian guns beyond the reach of Russian drones.

Steelberg’s business partner, Andy Baynes, had spent four years reporting to Steve Jobs at Apple. Together, the two set out to build Sceptre, a shell designed to fire from a standard NATO 155mm howitzer, and started raising money.

There’s a familiar story here about startup ambition discovering weapons, and the unease of turning that ambition towards killing. But the underlying idea is much older. Steelberg’s model calculated which military intervention would deliver the greatest effect for the money, literally the biggest bang for the buck. During the Second World War, physicist P.M.S. Blackett and his team, known as Blackett’s Circus, did much the same. They used mathematics to improve anti-aircraft guns, protect convoys, and hunt German submarines. AI makes the calculations faster and more accessible. The idea of treating war as an engineering problem is at least 80 years old.2

What’s changed is who gets to do the calculating. Blackett worked for the British state. Tiberius does it through a privately owned AI platform, funded by venture capital. Its business model is borrowed from Apple and the semiconductor industry: design the product yourself, outsource its manufacture, and avoid the cost of owning factories. Applied to weapons, this is what we might call fabless lethality. The two ideas reinforce each other: Tiberius owns the thinking, and the infrastructure through which other manufacturers turn that thinking into hardware.

Steelberg has a tidy way of putting the ambition. When you hit a hard problem, he says, you cross it, then build a bridge so the rest of the industry can follow. »It’s a toll bridge, obviously,« he explains. Others can pay a small fee to cross, »but it’s better than dying in the ditch«. The last phrase is almost too literal, given the business.

That doesn’t make the military case any less compelling. Longer-range artillery could let Ukrainian gunners operate beyond the reach of Russian drones, and Ukraine has every reason to want that advantage. Tiberius also has government partnerships, a reported $500 million UK collaboration,3 and investors with direct national-security interests. This isn’t a company operating outside the defence establishment. But its role goes beyond supplying weapons. It’s also developing the system that decides what weapons should be built.

The comparison with Blackett raises a question beyond whether the weapon works. Blackett’s researchers calculated inside a state-directed war effort. Their findings informed decisions made by military and political authorities outside the laboratory. At Tiberius, the questions of what the war needs and what the market will buy are answered inside the same company, by the same model. Blackett supplied information to those making military decisions. Tiberius is making those decisions part of its business.

Robert Oppenheimer needed a state, an extraordinary concentration of scientific talent, and years, to build a new class of weapon.4 Tiberius needed two technology executives, an AI platform, and a Series A. The comparison isn’t about the weapons. It’s about how far the price of admission to this kind of problem has fallen.

Steelberg says the decision began with housing Ukrainian refugees in his French château, an experience that left him heartbroken. He’s equally direct about the commercial ambition: the business is meant to make money. Neither claim cancels out the other. A company can start out wanting to help Ukraine and end up with a commercial interest in the weapons it develops.

Blackett’s Circus helped the military make better use of the weapons it had. Tiberius uses AI to identify what weapons should be developed in the first place. Its model can calculate which weapon is most effective, but that’s only part of the decision. Someone still has to decide what a country is trying to achieve – and what it is prepared to spend, risk, and kill to achieve it.

1 Isabella Ward (2026) »AI Told These Ex-Google Guys How to Build a Billion-Dollar Weapons Business«. WIRED, 17 November 2026. https://www.wired.com/story/ai-told-these-ex-google-guys-how-to-build-a-billion-dollar-weapons-business/

2 INFORMS (2026) »Patrick M. S. Blackett«. INFORMS History of Operations Research Excellence, Biographical Profiles. https://ez.informs.org/Explore/History-of-O.R.-Excellence/Biographical-Profiles/Blackett-Patrick-M.-S; Royal Air Force (2026) »Operational Research in the Royal Air Force«. RAF Air Historical Branch, AP3368. https://www.raf.mod.uk/what-we-do/our-history/air-historical-branch/second-world-war-thematic-studies/ap3368-operational-research-in-the-royal-air-force/

3 Army Recognition (2026) »Tiberius Advances Sceptre 155mm Ramjet Munition Into Allied Production Under $500 Million Deal«. Army Recognition, 1 October 2026. https://www.armyrecognition.com/news/army-news/2026/tiberius-advances-sceptre-155mm-ramjet-munition-into-allied-production-under-500-million-deal

4 Richard Rhodes (1986) The Making of the Atomic Bomb. New York: Simon & Schuster.


Previous editions

See the full archive →

Subscribe

Get it daily, or just the weekly roundup.