The Death of the Defense Prime: Inside the Pentagon's Plan to Starve Lockheed, Boeing, and Raytheon

On 17 June 2026, the Secretary of the Army put the defense-industrial establishment on a public clock. Speaking on the technology-business broadcast TBPN, Daniel Driscoll framed the next two years as a make-or-break window for the established primes — Lockheed Martin, Boeing, Raytheon — and made the bet explicit. "I will measure it as success," he said, "if in the next two years, one of the primes is no longer in business and the rest of them have all gotten stronger." It was a rare, unscripted shot across the bow from a sitting service secretary, delivered on a show that bills itself as covering the new economy.

The Army's pivot rests on four pillars that, taken together, amount to a quiet demolition of the post-Cold War procurement model. The service is cutting Congress-mandated legacy systems, redirecting the savings into drones, autonomy, and data infrastructure, and reclaiming the right to 3D-print its own parts — explicitly banning contractual restrictions on self-repair. Roughly 1,000 headquarters staff are being pushed back into combat formations. General George, the Army Chief of Staff, described a parallel shift in kit: drones that receive software updates between brigades, built on modular, open-architecture lines rather than the frozen-platform procurements of the past.

The data layer is the connective tissue. Driscoll pointed to Army Recruiting Command's migration off proprietary software and onto Salesforce out-of-the-box as the template — buy commodity enterprise tools, stop building bespoke. The pitch to Silicon Valley was direct. "We are inviting you in," Driscoll said. "We are inviting that community to come help us." Founders were invited to test hardware at combat training centers, an offer that would have been unthinkable a decade ago, when the primes jealously guarded evaluation slots as a procurement moat.

The math problem the primes can't solve

The underlying grievance is a numbers problem, and Driscoll laid it out bluntly. A Robotic Combat Vehicle runs the Army roughly $3 million per copy; an $800 drone can destroy one. "The math doesn't work," he said. "We're one of the wealthiest nations in the history of the world." For a service that spent two decades procuring exquisite, low-rate platforms optimised for counterinsurgency, the cost-exchange arithmetic of a near-peer drone fight is unforgivable. Legacy platforms cannot be retro-fitted into a procurement model that prices attritable mass at consumer-electronics levels.

That realisation, more than any ideological preference for startups, is what is driving the cultural shift. The Army hit its 12-month retention goals in 6 months; recruiting is up across nearly every category and geography. The force is healthier, and now wants a supply chain to match.

Anduril's M&A template and the $3–400 million ceiling

The most consequential private-sector response is Anduril's. Adam Porter Price, the company's head of M&A, walked through the firm's Class acquisition — a maker of ruggedized computers embedded in JLTV and Lattice deployments — and used the moment to lay out a thesis. Anduril targets companies for 9 to 12 months before approaching, pays market multiples, and frames every deal as a 5–10x revenue opportunity over 3–4 years. The operating principle is blunt: "We don't buy companies that are for sale."

Porter Price also offered a sober number for the rest of venture-backed defense tech. The ceiling for a VC-funded defense exit, he said, is roughly $3–400 million. Above that, boards apply rational valuation math and walk. He cited Boeing's roughly $300 million acquisition of Liquid Robotics as the proof point: a defense-adjacent platform sold to a strategic at a price that leaves venture math flat.

The structural reason matters. Delian Asparouhov of Founders Fund argued on the same broadcast that defense-tech M&A is fundamentally constrained: there is no "uncontrolled exponential growth" that forces incumbents into defensive buying, the way AI applications have triggered a land grab at OpenAI's reported $3 billion Windsurf deal. Capella Space, Asparouhov noted, sold for roughly $320 million — almost exactly its $315 million preferred stack — after a propulsion failure, a Rocket Lab launch loss, and the CEO's exit. Heavily-funded defense startups that build a science project rather than a product face a flat outcome, not a windfall.

For Anduril, the model works because the company is itself a strategic acquirer with Pentagon pull, not a portfolio company waiting for a trade sale. For everyone else in the defense-VC stack, Porter Price's number is a warning.

Silicon Valley as the new supply chain

The Prime's loss, in this telling, is Silicon Valley's gain — and the Army is choreographing the handoff. By inviting founders to test gear at combat training centers and pointing to Salesforce-style commodity enterprise buys, Driscoll is doing two things at once: lowering the barrier to entry for non-traditional vendors, and forcing the primes to compete on price and modularity rather than on access. Harley Finkelstein, Shopify's president and a separate guest on the same broadcast, offered a useful parallel. Shopify — at roughly 12 percent of US e-commerce share against Amazon's 40 percent — has built a second-place alternative not by out-Amazoning Amazon, but by owning the operating layer beneath thousands of merchants, with a developer ecosystem of 16,000 apps and over $1 billion paid in revshare in 2024.

The pattern translates. The Army is signalling it wants to be a platform, not a customer of finished platforms. Shop Pay's $22 billion in first-quarter GMV, up 57 percent year over year, was Finkelstein's way of illustrating what infrastructure scale looks like when you stop trying to be the store and start trying to be the rails. The Pentagon's rails, in this analogy, are open-architecture data, additive manufacturing, and software-updatable drones.

Counterpoint: the primes still own the production line

The skeptics have a case. The primes are not charities; they are incumbent producers with deep Congressional relationships, classified programme access, and a workforce in 44 states. Driscoll's two-year clock is, in practice, a 24-month political window inside an acquisition system that runs on five-year budget cycles and ten-year platform lifecycles. The Salesforce-for-recruiting analogy is encouraging, but recruiting is a back-office function; the harder question is whether the Army can replicate that speed in missile production, submarine welding, or nuclear-certified avionics. Asparouhov's own Capella anecdote is a useful corrective: most VC-funded defense plays have not produced Anduril-style outcomes, and the $3–400 million ceiling suggests the math is harder than the rhetoric implies.

There is also the question of who absorbs the political cost when a prime collapses. Defence industrial base employment is a bipartisan concern; the constituencies that lose when Lockheed shrinks a line in Marietta or Raytheon closes a plant in Tucson are not the same constituencies funding a Salesforce migration in Arlington. The Army can invite Silicon Valley in; Congress decides who leaves.

Stakes: a $200 billion question, two ways

The broader market context, aired in the same broadcast, sharpened the stakes. A TBPN host put Google's search-revenue cash cow at over $200 billion and growing — the kind of franchise value that explains why incumbents fight disruption. The defense primes occupy a comparable position in their own market: indispensable to the state, buttressed by regulation, and enormously profitable on the platforms that exist. The Driscoll bet is that the next two years will demonstrate which of those platforms survive software-defined warfare, and which are absorbed, retired, or — in his own framing — allowed to fail.

If he is right, the 250-year-old US Army will look more like a Shopify than a Sears by 2028: a platform orchestrating thousands of vendors on open standards, buying mass and software updates instead of monuments. If he is wrong, the primes endure, the next acquisition cycle stretches into the 2030s, and the $3 million RCV-versus-$800-drone asymmetry becomes the signature procurement scandal of the decade. Either way, the bet is now public, the timer is running, and the founders are taking meetings.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://www.youtube.com/watch?v=wOp99fhL_hc