Pull the Plug on Weak Products
The biggest mistake defense startups make
The real moat in defense is not just building a good product, it is killing weak ones early enough that the company still has the cash and focus to scale the winners. In this market, every self funded product bet ties up engineers, test time, and working capital long before a large program exists. Anduril worked because it moved from fast prototypes to paid trials to scaled programs, while keeping a tight grip on which products could become repeatable lines of business rather than endless custom projects.
-
Anduril did not start by trying to win a giant prime contract. It started with small paid work, then a $12.5M Marine Corps contract, then larger programs. That staged path made it easier to see which products were earning trust and which were not worth feeding with more internal R&D dollars.
-
This discipline matters because defense startups are usually self funding development to avoid cost plus economics. That can produce software like margins, but only if products become repeatable commercial or non developmental items. If a product turns into bespoke engineering for one buyer, the company gets services economics instead.
-
The comparison set shows why most companies should run narrower playbooks. Forterra shut down 17 legacy contracts to focus on one autonomy stack across defense and commercial vehicles. Shield AI is focused on autonomy software for aircraft. Saildrone built around ocean drones. Few teams can support many product lines at once.
The next wave of defense winners will look less like sprawling labs and more like ruthless capital allocators. The companies that matter will keep one core product base, extend it into adjacent programs, and cut experiments before they harden into expensive distractions. As budgets tighten and more startups crowd into defense, that kill discipline becomes a bigger advantage than invention alone.