The Epirus Playbook

COMMANDER · ISSUE 18 · October 19, 2026 · Istanbul

On a test range at Camp Atterbury, Indiana, in August 2025, a squat, panel-faced machine the size of a shipping container pointed itself at the sky and waited. Sixty-one drones went up. Sixty-one came down. One of the engagements took out a swarm of forty-nine aircraft in a single pulse — no missiles fired, no rounds spent, nothing in the air but an invisible wall of energy. The machine is called Leonidas, and the company that built it, Epirus, was founded in 2018 to answer a question almost nobody in air defense was asking yet: what do you do when the thing attacking you costs a thousand dollars and there are hundreds of them?

By March 2025 that bet had pulled in a $250 million Series D and more than half a billion dollars in total funding, a strategic investment from General Dynamics, and a string of U.S. Army contracts. Epirus did not get there by building a better missile. It got there by refusing to play the missile's game at all — by attacking the one number every legacy air-defense system quietly loses on: the cost of the shot.

The business behind the hardware. Not what they build, but how they built it.

Everyone else in defense media covers the weapon — the range, the power output, the physics of frying a drone's circuits from a distance. Almost nobody covers the company: how a startup priced a new category, funded it before the customer would, and climbed the acquisition ladder rung by rung until the Army was buying production units. That gap is why Commander exists. Here is the Epirus playbook — the moves an operator can steal.

Epirus was founded in September 2018 by a group including Nathan Mintz, Dr. Bo Marr, and Joe Lonsdale, incubated through Lonsdale's 8VC. At that moment the U.S. air-defense budget was still built around aircraft and ballistic missiles — big, expensive, relatively rare targets met with big, expensive interceptors. Cheap attack drones were a hobbyist curiosity and a niche worry, not a line item. Epirus started building a counter-swarm microwave weapon anyway, years before proliferating drones turned into the defining security problem of the decade and every defense budget on earth scrambled to catch up.

Founding a company on a threat the market has not yet priced is the highest-variance bet in defense, and the highest-return. If the threat stays niche, you die early and quietly. If it arrives, you are the only vendor with three years of hardware and flight data while the incumbents are still writing requirements. Epirus did not predict the future better than the Pentagon; it committed capital and calendar to a specific answer before the demand curve moved, so that when the curve moved, it was already standing there.

Takeaway: The best time to build for a threat is before the market believes in it. You trade years of lonely risk for the position no fast-follower can buy back once the threat is obvious.

The pitch that built Epirus is not really about microwaves. It is about an exchange rate. Shooting down a $1,000 drone with a six- or seven-figure interceptor is a trade you lose even when you win — a few dozen cheap aircraft can bankrupt the magazine and the budget of anyone defending with kinetics. Leonidas inverts that arithmetic. A shot is a pulse of directed energy: pennies of electricity, an effectively unlimited magazine, and — because a microwave beam is a cone, not a bullet — many targets killed at once. Epirus reframed air defense from "one interceptor, one threat" to what its team calls a "1-to-many" problem.

That reframing is the whole wedge. When the incumbent solution loses money on every success, you don't have to build a marginally better version of it — you change the unit economics of the fight. Epirus sold defense buyers a spreadsheet before it sold them a box: cost-per-engagement, targets-per-shot, rounds-you-never-run-out-of. The hardware is impressive, but the argument that opened doors was financial.

Takeaway: When the standard solution loses money on every win, your product is the new cost curve, not the new gadget. Lead with the exchange rate; the buyer's own math becomes your sales pitch.

Directed-energy weapons have existed in the lab for decades and stayed there, because the old way of making them — vacuum tubes, klystrons, exotic one-off hardware — was fragile, power-hungry, and impossible to manufacture at scale. Epirus built Leonidas out of gallium-nitride solid-state amplifiers, the same class of semiconductor material that powers modern radar and telecom, coordinated by a software-defined power-management system the company calls SmartPower. It is a solid-state, digitally controlled machine with an open architecture, not a hand-tuned physics experiment.

The choice of technology base is a manufacturing decision disguised as an engineering one. Solid-state GaN can be cooled, packaged, upgraded in software, and — critically — produced on supply chains that already exist for commercial electronics. Epirus did not just make a directed-energy weapon that works; it made one that can be built on a production line and improved with a code push. In hardware, the component base you pick in year one decides whether you can ever scale in year six.

Takeaway: Pick the technology base that scales like a factory, not like a lab. The winning version of a hard product is usually the one built from components someone else already mass-produces.

Epirus raised like a software company to build like a defense one: a $17.4 million Series A in 2019, $70 million in 2020, $200 million in 2022 at a valuation near $1.4 billion, and the $250 million Series D in March 2025 — more than half a billion dollars in total. That capital paid to design, build, and demonstrate a working weapon before any formal program of record existed to buy it. When the Army was ready to move fast, Epirus already had hardware to hand it.

This is the inversion of the traditional defense model, where a contractor waits for a requirement, wins a cost-plus contract, and gets paid to develop slowly on the government's clock. By funding the hard, expensive proving phase with venture money, Epirus arrived at the government's door with a finished demonstration instead of a proposal — and stayed the owner of its own roadmap and intellectual property. Capital, spent ahead of the requirement, buys the one thing a cost-plus contract can't: being ready before the customer is.

Takeaway: In defense, private capital's job is to make you ready before the requirement is written. Show up with working hardware and you set the terms; show up with a proposal and the customer sets them.

In December 2023 Epirus brought in Andy Lowery — a retired Navy officer and technical founder — as CEO, putting an operator's credibility at the top of a company selling to operators. Then it did the thing most hardware startups get wrong: instead of trying to out-muscle the defense primes, it recruited one. General Dynamics Land Systems came in as a strategic investor and integration partner, the route to mounting Leonidas on the Army's Stryker vehicles and into the logistics chain a startup cannot build alone.

A startup can invent a category; it almost never has the balance sheet, the program relationships, or the sustainment footprint to field it at national scale. The primes have all three and rarely the appetite to invent. Epirus treated that asymmetry as a channel, not a threat — let the prime do integration and scale while Epirus owned the core technology and the roadmap. The lesson is not "avoid the incumbents." It is "decide which part of their machine you want to rent."

Takeaway: Don't try to beat the primes at their own game — recruit one as a channel. Own the novel core; rent their scale, relationships, and sustainment.

In a category the market does not yet believe in, the demonstration is the sales team. At Camp Atterbury in August 2025, Epirus ran a live-fire event in front of Army evaluators: 61 drones flown, 61 defeated, including a 49-aircraft swarm dropped in a single pulse. In separate testing, Leonidas took down a fiber-optic-controlled drone — the kind that is immune to radio jamming because it trails a physical data cable, and therefore beats the electronic-warfare tools the field has leaned on. Microwave energy burns the electronics regardless of how the drone is flown.

Each of those results converts a claim into a witnessed fact. "Unlimited magazine" is a slide until forty-nine drones fall at once in front of the people who write the checks. "Works against threats that beat jamming" is marketing until a fiber-optic drone drops on the range. Epirus understood that in an unproven category, skepticism is the real competitor, and the only thing that kills skepticism is a repeatable demonstration the buyer watches with their own eyes.

Takeaway: In a category nobody believes yet, your best salesperson is a public, repeatable demo. Turn every disputed claim into a fact the buyer witnessed — then do it again on command.

Epirus did not wait for one giant contract; it stacked small, fast ones into momentum. A roughly $66 million Rapid Capabilities Office award in 2023 funded four Leonidas prototypes. Those became four IFPC-HPM Generation I systems delivered to the Army in May 2024. A sensor-suite modification followed, then a Navy contract to build an expeditionary variant. In July 2025 came a $43.5 million award for two Generation II systems — more than double the range and thirty percent more power than Gen I. The Series D exists to hyperscale production for exactly this demand.

No single one of those awards is a program of record. Together, in sequence, they are the on-ramp to one. Each delivery earned the credibility to win the next, larger contract, and each generation folded in the lessons of the last. In defense procurement, the rapid-acquisition path rewards the vendor who can turn a prototype award into a delivered system fast enough to justify the follow-on — and Epirus optimized its whole company around that cadence.

Takeaway: Don't wait for the program of record — build it out of a chain of small, fast awards. Each delivery you actually ship is the argument for the next, bigger one.

Start building for the threat before the market prices it, and trade years of lonely risk for a position no fast-follower can buy back. Sell the exchange rate, not the machine: when the incumbent loses money on every win, your product is the new cost curve. Build on commercial silicon so the thing scales like a factory instead of a lab. Fund the expensive proving phase with private capital so you arrive at the customer's door with working hardware, not a proposal — and keep your roadmap. Hire the operator for credibility and recruit a prime as your channel to scale rather than fighting it. Prove everything in public, because in an unbelieved category a repeatable demo is your best salesperson. Then ladder a chain of small, fast awards into an unofficial program of record. The last archetype won by refusing to build the final product first; this one wins by refusing to fight on the enemy's cost curve.

This issue's company won by refusing to fight on the enemy's cost curve — it changed the price of the shot until the math flipped. The next archetype wins somewhere else entirely: not on unit economics, but on becoming the one supplier everyone else has to build around — the company that made itself a dependency rather than a product. No company names yet. Collect the set: each issue is one operating archetype, and this one — invert the cost curve, price the threat before the market does — sets up the archetype that wins by owning the chokepoint.

— COMMANDER

A new Playbook every Monday, the Weekly Transmission every Friday. If it earned its place in your inbox, forward it to one operator who's building something — that's how Commander spreads.