Mistaking Cultural Shift for Policy Change
The biggest mistake defense startups make
The bottleneck in defense has shifted from permission to execution. New startups are not entering a newly deregulated market, they are entering a market where the buying machinery still moves slowly, but engineers, operators, and investors now believe a venture backed company can survive that slog. That belief was built by Palantir, SpaceX, and then Anduril proving that a startup could win real programs without acting like a traditional cost plus contractor.
-
The practical rulebook is still familiar. Small contracts build trust, then expand into larger awards over years. In the interview, Scott Sanders describes Anduril starting with small paid work, then a $12.5 million Marine Corps commercial item contract, then eventually a billion dollar scale program. That is a sales motion and credibility ladder, not a policy shortcut.
-
What changed most was talent and capital formation. In 2017, defense tech was not a mainstream startup category and many VCs saw multiyear procurement, hardware, and non recurring revenue as disqualifying. By early 2024, Anduril had already scaled from an estimated $4.8 million of revenue in 2018 to $420 million in 2023, which turned cultural skepticism into proof that the category could compound.
-
This is why founders misread the market when they treat DoD as one giant customer waiting for better tech. It is a patchwork of micro markets, budget lines, and program managers. The companies that break through do not wait for rules to improve, they package products in forms the government already knows how to buy and line them up against an immediate mission need.
The next wave will look less like a sudden opening of Pentagon procurement and more like a growing class of companies trained by the first winners. As more operators leave Anduril, Palantir, SpaceX, and Forterra with firsthand knowledge of how to turn prototypes into programs, the culture shift should keep widening, even if the procurement plumbing stays mostly the same.