Silicon Valley Goes to War, and Brings a Pitch Deck
Valley of Death by Sharon Weinberger
Little, Brown and Company · September 2026
In 2013, about ten
scientists and strategists gathered in a Virginia home to play a game whose
premise sounds like the pitch for an airport thriller: could a rich man build
and sell nuclear weapons? The answer, reached with nothing more than
open-source research and shopping on Alibaba, was yes. It would take roughly a
billion dollars, about five years and four nominally unrelated companies spread
from Zurich to Kazakhstan to Dubai, and it would be hard for anyone to stop.
The Pentagon had commissioned the exercise, and by one participant's account
the people involved were told to scrub their computers afterward, lest the
study double as a how-to manual. Sharon Weinberger opens Valley of Death with
this war game, and it is a shrewd choice: a lurid, memorable image that carries
her argument before she has made it.
The argument is that the government's old monopoly on the machinery of war is passing to a small band of technologists and venture capitalists who have their own politics, their own balance sheets and a deep suspicion of the institutions they sell to. She calls the new arrangement the Pentagon-Silicon Complex, a deliberate echo of the military-industrial complex that Eisenhower warned about in 1961. The title borrows an industry cliché, the “valley of death” between a promising prototype and a paying military customer. The book's real subject is what happens now that the valley is being crossed. The numbers she offers are startling: by her account, venture capital poured nearly $50 billion into defense startups in 2025, more than double the previous year.
Weinberger, a veteran national security reporter who spent years at the Wall Street Journal and wrote an earlier book on DARPA, is at her best when she goes backward in time. The first chapter is a small marvel. It begins with a teenage mathematician recruited by the CIA in 1969 and arrives, improbably, at Steve Jobs, whose struggling NeXT computer company was kept afloat by spy-agency orders. The machines were expensive, elegant and almost unsellable to ordinary customers; the intelligence community bought them by the thousands, and a billionaire investor leaned on Jobs when he hesitated. When Apple bought NeXT in 1996, the operating system that came with it quietly shaped the company's future. The anecdote punctures one of Silicon Valley's favorite myths, that it is a civilian garden of garage tinkerers and counterculture dreamers, and it establishes the book's governing irony: the Valley was grown on Pentagon money long before it started pretending otherwise.
From there the book becomes a group portrait, and the cast is formidable. Elon Musk, who wanted to go to Mars and found that the Pentagon would help pay for the rockets. Peter Thiel, who told a Stanford class that competition is for losers and meant it as a business plan, not a provocation. Alex Karp, Thiel's partner in Palantir, a company named after the seeing stones of Tolkien and born from the failures of intelligence sharing that preceded September 11. Palmer Luckey, the virtual-reality prodigy turned weapons maker. Weinberger sketches them with a reporter's economy, usually through a telling scene instead of a character study, and she is careful to let their ambitions speak for themselves. Thiel's monopoly gospel, in particular, reads very differently once you notice how closely SpaceX's grip on military launches, and Palantir's own ambitions, resemble it.
The book is sharpest when it deflates its subjects. Take Project Maven, the Pentagon's early attempt to use artificial intelligence on drone footage, remembered mostly for the employee revolt that drove Google away from it. Weinberger shows that the rebellion mattered less than anyone thought. The system's first field test went badly enough that, by a Palantir executive's own telling, it struggled to tell people from sheep. Its architects concede that the Pentagon lacked the basic digital plumbing, data pipelines and cloud infrastructure that any serious AI program needs. What Maven really did, she argues, was open the floodgates, because a wave of companies saw that working with the military was now respectable and lucrative, particularly as the Chinese market closed to them. It is a nicely counterintuitive account, and it is typical of her method: take the official story, find the people who lived it, and let the gap between claim and reality do the work.
Politics enters in force in the middle chapters, and Weinberger handles it with unusual restraint. Thiel's endorsement of Trump at the 2016 Republican convention was a shock in a Valley that largely recoiled from the candidate. The fight over JEDI, a $10 billion Pentagon cloud contract, turned into a proxy for Trump's feud with Jeff Bezos, and the president needed a moment just to identify the contract when a reporter asked him about it. Palantir went public in 2020 and was worth more than $20 billion by the end of its first day of trading. Then Biden's election, perversely, gave the industry something Trump could not: legitimacy among the left-leaning engineers whose labor it needed. The book is good on the way a movement that began as an embarrassment became, within a few years, a career path with a fan base.
The war chapters are where the title's anxiety turns concrete. Russia's invasion of Ukraine made weapons-building fashionable, and Starlink terminals, commercial satellite imagery and Amazon's cloud became instruments of a real war. There is a wonderful, absurd image of Palantir's chief executive doing tai chi near the Polish border with Ukraine, drawing suspicious looks from guards. Weinberger turns to a harder question in her conversation with Luckey about Musk's control over Starlink in Ukraine: if the company's argument is that elected governments, not executives, should set foreign policy, what happens when elections put those executives in power? Luckey never quite answers, and her refusal to let him off is the best moment of interviewing in the book. The same instinct drives her account of Gaza, where Palantir announced that it stood with Israel, an unusual political stance for a defense contractor, and she asks plainly why a company would choose to.
Trump's second victory is where the book arrives at its most chilling formulation. Karp appears at the Reagan National Defense Forum in December 2024, practically giddy, attacking Berkeley, the United Nations and Democrats, and promising that the company's customers will win everything. Musk answered the speech online with a single word, “Based.” Weinberger lets the scene sit, observing that no one in the room, including Democratic lawmakers, contradicted him, and she ends the chapter with a sentence that earns its bluntness: the war bros were no longer just lobbying the government, they were in it. The next chapter opens on the tech leaders' dinner with Trump in September 2025, a ceremony of flattery that she rescues from caricature by quoting a venture capitalist's dry remark that corporate power and state power have simply replaced the royal court.
For all its virtues, Valley of Death has real weaknesses, and they are mostly structural. The book is episodic by design, a chain of vivid scenes held together by theme more than by narrative momentum, and a reader without a specialist's familiarity with defense contracting may drown in the names of companies, programs, offices and investors. Palantir, Karp and the Thiel network occupy so much of the foreground that other actors, including the traditional primes the new firms claim to displace, recede into the background. And the book's central metaphor is slipperier than it first appears. Weinberger concedes that these founders probably won't build nuclear weapons; the real worry is autonomous drones, AI-assisted targeting and software that quietly shapes decisions. The billionaire's bomb is an arresting frame, but it promises a danger the book then relocates, and the shift is not always clean.
That said, the book's most admirable quality is its refusal to be a polemic. Weinberger is wary of the concentration of power in a few hands, but she is equally wary of the industry's critics who assume these companies are competent. Her deepest skepticism is reserved for the proposition that the new weapons work. A venture investor tells her, in effect, that it does not matter to him whether what he funds works. A drone company's chief executive explains, over wine at a Northern California vineyard, that he wants to own the first twelve choices on Amazon's list so that customers feel they are choosing freely. The firm, valued at more than $3 billion against modest revenue, appeared in a federal database to hold one government contract worth $167,000. A General Atomics spokesperson went so far as to call Anduril the Theranos of defense, a comparison Weinberger notes is unfair, since no one has alleged fraud, but she lingers on it because the rhetoric, hype and valuation logic are familiar.
The epilogue turns to the scenario that justifies the whole enterprise: a Chinese invasion of Taiwan. Weinberger describes tabletop war games in which the United States and its allies usually win but at a terrible price, with two aircraft carriers and thousands of sailors lost. She sets that grim arithmetic beside a Palantir commercial that depicts a nearly bloodless war fought by drones and a commander conducting them like an orchestra, a vision of warfare that, she points out, does not yet exist and that Palantir is not building. Investors, in her phrase, were betting on vibes. Michael Burry's decision to short the stock and the company's decline of about a quarter from its late-2025 high, as of March 2026, serve as a reminder that the market's enthusiasm may be running ahead of the evidence. The risk, she suggests, cuts both ways: a bust could drag down a whole class of startups just as the Pentagon has learned to depend on them.
She closes with a conversation with Will Roper, the official behind Maven, who urges her to end on the hundreds of companies caravanning across the valley of death and the possibility that the experiment fails if no one welcomes them on the other side. Weinberger chooses a slightly different ending. The danger, she says, is not that Silicon Valley will fail to deliver, but that a few firms will succeed so thoroughly that they displace the old primes and carry their founders' convictions into the heart of national security. She returns to Eisenhower, who understood the need for the industrial development he was warning about and urged the country not to fail to grasp its grave implications. It is a restrained and fitting conclusion, and it leaves the reader with a question rather than a verdict.
Valley of Death is not a perfect book. It is crowded, occasionally repetitive and more convincing as reportage than as prophecy. But it asks the right question at the right moment, and it asks it with a rare combination of access, skepticism and fairness. The pleasures here are the pleasures of good journalism: the unexpected anecdote, the telling detail, the interview in which the subject says slightly more than he meant to. For anyone trying to understand who will build, sell and ultimately control the next generation of weapons, and what that might do to a democracy that has long assumed it holds the reins, this is essential reading. Read it for the scenes. Stay for the unease.
