The Shield AI Playbook

In 2015, when the two brothers behind Shield AI went looking for money to build autonomy for defense, most investors wouldn't touch it. One offer did come in — around $5 million, on the condition that they drop the defense mission and point the technology at something more comfortable. They said no. The first money they actually took was an $800,000 seed, roughly six times smaller, from someone who believed in the thing they wanted to build.

A decade later, that company raised $2 billion in a single round at a $12.7 billion valuation, and its software has flown twenty-six different classes of vehicle — from a hand-thrown quadcopter to an F-16. The interesting part isn't the size. It's that almost every hard decision Shield AI made points the same direction: the aircraft was never the product. The software was.

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

Most coverage of a company like this reaches for the hardware — the drone, the jet, the spec sheet. That's the least instructive part. The story worth stealing is how a founder refused the easy money, decided which business he was actually in, and then bought, built, and licensed his way into owning a layer instead of a product. That gap is why Commander exists. Here are the seven moves.

The $5 million offer would have bought a longer runway and a completely different, smaller company — one pointed away from the market its founders actually understood. The $800,000 they took instead came with conviction attached rather than conditions. That single refusal set the trajectory: it kept the company aimed at the one wedge where its founders had an unfair advantage, instead of drifting toward whatever was easiest to fund that quarter.

Early capital feels like survival, so founders take the version with the fewest questions asked. But money that redirects the mission is the most expensive money on the table, because you pay for it with the company you were supposed to build.

Takeaway: The wrong capital costs more than no capital. Money that bends the mission is a loan against the company you actually wanted.

Shield AI's first product, Nova, was a small drone soldiers could throw into a building. But the drone was the delivery vehicle, not the product. What mattered was that it flew itself — no GPS, no operator, no comms — which in 2018 made it the first AI-piloted aircraft used in real combat conditions. From the start the company knew which business it was in: it makes an AI pilot, called Hivemind, and occasionally wraps it in hardware.

Everyone else in the drone market was selling airframes and treating the autonomy as a feature. Shield AI inverted the stack. The plastic and carbon fiber were interchangeable; the brain was the asset. Get that call right and every later decision — what to build, what to buy, what to license — gets easier, because you know what you're protecting.

Takeaway: Decide which business you're actually in. If the software is the scarce thing, the hardware is just the place it runs — price and defend accordingly.

In 2021 Shield AI made two acquisitions in quick succession: Heron Systems, the team that had just won DARPA's AI-versus-human flying contest, and Martin UAV, which brought a proven runway-independent aircraft — the V-BAT — and a factory to build it. In 2024 it added Sentient Vision for maritime sensing. Each deal folded in a capability that would have taken years to grow from scratch: a world-class autonomy team, an airframe with 13-plus hours of endurance that needs no runway, an optical payload.

Building all of that in-house would have been cheaper on a spreadsheet and years too slow in reality. In a category being decided in real time, acquisition is a speed tool, not a vanity one. You're not buying revenue — you're buying finished time.

Takeaway: When the clock is the constraint, buy the team, the airframe, or the factory. Growing it yourself is the expensive option measured in the currency that matters — time.

Hivemind isn't tied to a single aircraft. It has now piloted twenty-six classes of vehicle — quadcopters, jet-powered drones, helicopters, uncrewed boats, ground vehicles, and F-16s. That generality is the entire strategy. A company that builds a great drone has a product. A company that builds a pilot that flies anything has a platform, and platforms compound: every new airframe Hivemind masters makes the next one cheaper, and the value accrues to the software layer rather than the metal.

It's the harder thing to build and the more valuable thing to own. A point solution wins one program; a horizontal layer wins the ones that haven't been written yet.

Takeaway: Build the horizontal layer, not the point solution. A brain that generalizes across bodies beats any single body it lives in.

Most companies lock their best technology inside their own product. Shield AI does the opposite. It sells Hivemind as a stack — "Enterprise" — that other manufacturers can put on their own aircraft, and separately offers full turnkey systems for buyers who want the whole package. That's a deliberate choice to be the standard rather than just another competitor with a drone.

It's the difference between selling one phone and being the operating system every phone runs. When your software is the scarce asset, the widest possible install base beats a walled garden — you'd rather be the layer everyone builds on than the product everyone compares.

Takeaway: If the software is the moat, put it on everyone's hardware. Being the layer beats being one of the products sitting on top of it.

Shield AI's numbers move in steps, not slopes. Roughly $267 million of revenue in 2024. A $240 million round at a $5.3 billion valuation in early 2025. Then, about a year later, $2 billion raised — $1.5 billion in equity plus half a billion in preferred — at $12.7 billion, led by Advent International with JPMorgan co-leading. The valuation more than doubled in roughly thirteen months.

That isn't vanity; it's positioning. Autonomy for defense is being standardized right now, and the round funds the bet that matters: Hivemind was selected for the U.S. Air Force's Collaborative Combat Aircraft program and is flying on a next-generation uncrewed jet. You raise at that scale to be the default answer when the category finally sets.

Takeaway: When a market is still forming, capital buys default status. Raise to be the obvious answer before the question is settled — not after.

Running a company at a million dollars of revenue and running one at several hundred million are different jobs that happen to share a title. Shield AI treated that as a fact rather than a threat: the founders brought in a chief executive who had already run public companies at scale — Gary Steele, formerly of Splunk and Proofpoint — and moved themselves into the seats only founders can hold, product and mission.

Founders often read "bring in a CEO" as losing the company. Done on purpose and early, it's the opposite. It's refusing to let your own job description become the ceiling on what the company can become.

Takeaway: The founder's job changes as the company scales. Hiring the operator the next stage needs is a strength move, not a surrender.

[ BRANDED "BY THE NUMBERS" CARD — drop in the house stat card here, same template as issues 05–06. ]

Shield AI's answer to "where's the moat?" was: the software. Next issue flips the board. We take apart a company betting the real bottleneck in defense isn't the code or the airframe — it's the factory that can actually build hardware at volume — and that owning the least glamorous layer, the machined metal nobody wants to touch, is the opening everyone else is leaving open. The software company, mirrored by the metal-benders.

Commander is the business behind the hardware. Not what they build — how they built it.

New transmission every two weeks. If it earned its place in your inbox, forward it to one operator who's building something — that's how Commander spreads.

COMMANDER · ISSUE 11 · commander.media · Istanbul