An upsurge in defense expenditure is nudging technology firms away from catering to consumers and steering them back toward chasing government deals.
Valley of Death: How Big Tech Is Fueling the Future of War, by Sharon Weinberger, Little, Brown and Company, 320 pages, $32
Rumors suggest that Donald Trump consulted a chatbot before taking action in Venezuela. After an artificial intelligence purportedly warned that Venezuelans would celebrate the overthrow of their regime, the president allegedly left the operation with the impression that Grok was ingenious, Time reports, citing conversations with Trump and his advisers.
Only a few months afterward, another AI nearly stoked a fresh round of conflict. With the aid of a chatbot, a member of U.S. Special Operations Command drafted a briefing about a Chinese vessel delivering nuclear-kit components to Iran. The U.S. military almost attacked the ship before realizing the chatbot had misled the official, CNN reports.
Killer robots have long inhabited the popular imagination, but fewer people anticipated automated yes-men whispering in the ears of military leaders. Even odder is that these advisers are a consumer product anyone can access. Grok counted 117 million monthly users as of March 2026, according to SpaceX corporate filings, and what the U.S. Special Operations Command possesses is “mostly just copies of the commercial stuff wearing lipstick,” an official told CNN. Last year, the U.S. armed forces signed $200 million deals with OpenAI, Anthropic, Google, and xAI (SpaceX’s entity) to blend their systems into military databases.
“The most plausible use of AI models” by the military is to “identify possible targets that could later be destroyed,” Weinberger writes in her new book, Valley of Death. It mirrors the quandary she notes about Amazon offering cloud computing for the U.S. government: “A company that has years of data on your purchases—from books to medications—also supplies remote servers for intelligence agencies.”
Weinberger, a journalist who has covered the weapons industry for more than twenty years, argues that commercial tech firms and the military–industrial complex have profoundly reshaped one another. Instead of public corporations focused on weapons, the Pentagon hands contracts to billionaires who straddle the commercial and federal markets. Tech founders mold their strategies to lock in taxpayer-funded contracts and state-backed monopolies. Investors describe themselves as providers of “patriotic capital.” Government agencies rush to back promising technologies. The devices we use daily are being molded around military needs.
“Every drone company I had five years ago was a drone-delivery outfit trying to sell to Walmart, and today they’re weapon-first, with 90 percent of revenue coming from the military,” venture capitalist Clayton Williams tells Weinberger. Anduril founder Palmer Luckey, who built virtual-reality headsets for video games before turning to weapons development, predicts to Weinberger that “you’ll see augmented reality on every soldier’s head before you see it on every consumer’s head.”
Silicon Valley has never been entirely severed from state violence. The semiconductor makers that positioned San Francisco at the electronic frontier mostly supplied parts for military hardware. Larry Ellison, Oracle’s owner who has lately recast himself as a media magnate, entered the IT field by building a database for the CIA in the 1970s. And if a gap once existed between the countercultural hackers and the disciplined national security establishment, the CIA helped bridge it by rescuing Apple founder Steve Jobs. Jobs styled himself as fighting Big Brother and stubbornly refused to “do business for the fucking federal government,” yet his company NeXT would have collapsed in 1989 if the agency hadn’t purchased thousands of its computers, which private industry deemed expensive and impractical.
The CIA used those machines for satellite-imagery analysis, leveraging graphics tech created by Pixar. A decade later, the agency established a venture capital firm called In-Q-Tel to fund promising tech startups.
Among In-Q-Tel’s early bets was Palantir, a data-processing enterprise. Like Jobs, its founder Peter Thiel carried libertarian leanings in his early days. Thiel toldWired in 2001 that his best-known company, PayPal, would accelerate “the ability to move money fluidly” and contribute to “the erosion of the nation-state.” Less than two weeks after that issue appeared, Al Qaeda attacked the United States and Thiel redefined himself. In 2007, he argued that “the secret coordination of the world’s intelligence services” would be “the decisive path to a truly global pax Americana.”
A few years later, future OpenAI founder Sam Altman invited Thiel to deliver a guest lecture in a class Altman taught. “Competition is for losers,” Thiel preached, urging the class to seek monopoly instead.
Although Thiel aligns with the Republican Party, it was a Democratic administration that helped his ambitions take off. President Barack Obama began bolstering troops in Afghanistan in 2009, and Palantir pledged to help quell the Afghan insurgency by integrating diverse sources of surveillance data (a technology derived from PayPal’s fraud-detection methods). The firm provided U.S. troops with free access to its database software and even deployed its engineers to the country in a bid to render Palantir indispensable to the war. While Obama’s surge did not settle Afghanistan, Thiel’s gambit succeeded in winning him business.
Afghanistan also gave rise to a critical element of the AI revolution: image recognition. The U.S. military deployed Gorgon Stare, a drone-based camera system capable of surveilling an entire city, over Afghan terrain. (Weinberger notes that Gorgon Stare drew inspiration from the 1998 film Enemy of the State.) The system produced an overwhelming volume of footage for human analysts, so the military launched Project Maven, enlisting civilian AI firms to enhance image-recognition capabilities. When Google employees objected to participating, Palantir stepped in to secure the contract. Maven evolved into a broad battlefield-surveillance network, aiding the withdrawal of U.S. troops from Afghanistan and contributing to the 2026 aerial massacre of schoolchildren in Minab, Iran.
OpenAI and Anthropic, the leading AI research labs, originally vowed not to enter the business of killing people. Weinberger calls this “a distinction without a difference,” since they do not exclude helping to compile target lists. In 2024, OpenAI announced a collaboration with Anduril on target detection, and Anthropic formed a partnership with Palantir on “classified environments.” As Weinberger notes, these collaborations could turn ordinary queries into training data for autonomous weapons.
Valley of Death hints at a fundamental tension between how tech moguls speak and what they actually do. There is a near-orthodoxy in the tech world that the United States must be ready for war with China by 2027. Yet Silicon Valley’s comfort zone clearly lies with counterinsurgencies rather than battles between peer powers. Its technologies seem better suited to imposing control from above than to facing a peer adversary on equal terms.
The Russian invasion of Ukraine served as a dry run for great-power conflict. Although American tech executives tried to make a show of their help to the Ukrainian war effort, Palantir’s and Anduril’s offerings stood up poorly against locally developed Ukrainian systems. “It’s almost impossible to craft something that works without being intimately connected to the problem space,” Ukrainian technologist Andrey Liscovich told Weinberger, arguing that U.S. firms lose their nimbleness when they operate abroad. And the real challenge for Ukraine isn’t about making weapons smarter; it’s about producing them cheaply at scale. In short, great-power warfare is an issue of physical bottlenecks, not raw intelligence.
The industry’s eager involvement with Israel after October 7, 2023, which Weinberger presents as a deviation from its otherwise China-focused agenda, makes more sense as a return to the mean. The U.S. focus swung back toward countering low-tech guerrillas and managing unruly populations. That remained Silicon Valley’s comfort zone. While Luckey of Anduril continues to insist that China remains the top priority, he pitches a technology for dealing with tunnels in Gaza. He even hints that tomorrow’s competition with China could involve proxy wars similar to today, naturally.
In the conflict with Iran, Weinberger notes that there is still little to show that startup technologies are being deployed by the U.S. military (aside from Palantir’s Maven database and SpaceX’s satellite communications). And that may suit Silicon Valley just fine. With so much defense outlay aimed at securing tomorrow’s technologies, startup aerospace contractors can still reach billions in revenue without a single aircraft taking to the skies, a venture capitalist tells Weinberger. “Silicon Valley’s ‘fake it till you make it’ approach works until people die,” according to former CIA officer James Acuna, and that sentiment, she suggests, is not a universal concern within the industry.
Weinberger ponders whether entrusting power to private businessmen poses a threat to democracy. A perhaps more pertinent question is whether it helps or harms free-market dynamics. As Silicon Valley becomes increasingly insulated from the imperative to earn profits from customers, it shifts from technologies that enrich life and liberty to those that seek to undermine both.
