Valley of Death

The phrase in the title is one every capability developer knows by heart. The “valley of death” is the gap where a promising prototype goes to die — the stretch between a successful demonstration and a program of record with real money behind it. Sharon Weinberger borrows the term for a larger argument: that a new class of venture-backed technology companies has decided to cross that valley on its own dime, and in doing so is quietly rewriting who owns the future of American military power.

What the Book Argues

Weinberger, a national-security journalist of twenty-five years' standing, calls the result the “Pentagon–Silicon complex” and argues it represents the largest shift in the defense industrial base since the end of the Second World War. Where the old arrangement had the government define a requirement, fund the research, and own the result, the new firms — Palantir, Anduril, SpaceX, and a crowd of younger imitators — fund their own development, keep the intellectual property, and sell finished capability back to the government as a product. Power, in her telling, has migrated from the elected officials who write the checks to the unelected founders who own the code.

The book is at its best tracing what that inversion means in practice: billion-dollar contracts flowing to companies that build first and sell second, a defense market increasingly shaped by figures like Elon Musk, Peter Thiel, and Palantir's Alexander Karp, and a set of accountability questions the acquisition system was never designed to answer. She runs the thread out to its hard cases — a Taiwan contingency, semiconductor supply chains, AI woven into the targeting cycle — and asks who is actually in charge when the capability belongs to a vendor.

Where It Holds Up

Weinberger's central observation is correct, and it matters more than the Silicon Valley palace intrigue she wraps it in. For most of my career the government was the integrator: it held the requirement, the money, and the technical baseline, and industry built to that baseline. What she documents is the baseline moving inside the companies. When a vendor self-funds the R&D and retains the IP, the government stops being the architect and becomes a subscriber — and a subscriber has far less leverage than an owner.

Crossing the valley of death with private capital solves a real problem — the government's is too slow — but it quietly transfers the thing that used to give the government control: ownership of the capability itself.

Her reporting is the book's engine. Weinberger has the access to put executives on the record, and the resulting quotes are genuinely revealing about how these founders see the state — sometimes as a customer, sometimes as an obstacle, rarely as a partner of equals. For anyone on the government side of the table, that view from across it is worth the price of the book.

Where It Falls Short

Two cautions. First, the book is sharper on capital and ownership than on whether the technology actually works. It is appropriately skeptical of the marketing — the “kill chain” language, the sweeping claims about what these systems did in Ukraine or Gaza — and rightly treats some of it as exaggerated. But skepticism is not the same as measurement. The questions a practitioner most needs answered — how these tools perform under operational conditions, what their error rates are, what “human on the loop” means when the loop runs at machine speed — get less rigor than the boardroom drama.

Second, like most books built on access to founders, it drifts toward the personalities. Musk, Thiel, and Karp make for vivid copy, and a twenty-two-year-old weapons-company founder makes for a striking anecdote. But the deep story here is structural — contracting authorities, IP terms, the erosion of the government's technical workforce — and structure is less photogenic than the men. The reader has to supply some of the connective tissue the narrative skips.

What Leaders Can Use

Three takeaways that outlast the news cycle. First, read your terms, not the demo: the demonstration proves the thing can work once; the contract decides who owns it, who can fix it, and who you depend on for the next decade. Second, the government's real deficit is speed, and until the acquisition system can cross the valley of death at commercial tempo, private capital will keep crossing it first and setting the terms — which makes reforming that system a strategic issue, not a paperwork one. Third, if you sit on the buying side, invest in the technical competence to be a smart customer; you cannot govern what you cannot evaluate.