How Silicon Valley Swallowed the War Business: The Myths and Risks of Defense Innovation

Silicon Valley is redefining the underlying logic of the U.S. defense industry. From Palantir's data analytics platform to Anduril's autonomous weapons systems, a wave of venture-capital-driven startups is breaking the decades-long monopoly of traditional contractors like Lockheed Martin and Northrop Grumman. But this transformation is far from a simple narrative of "innovation defeating bureaucracy"—it also means that war-making decisions are shifting from elected governments to private tech giants, which face almost none of the constraints and oversight that the traditional defense industry does.

The The Gray Area podcast recently spoke with investigative journalist Sharon Weinberger, whose new book Valley of Death systematically traces the tangled history of the relationship between Silicon Valley and the Pentagon. This article distills the core contradictions and potential risks behind this "defense tech revolution" based on that interview.

From "Government Pays" to "VC-Driven": A Fundamental Shift in Defense Procurement

Weinberger first clarifies a common misconception in the interview: private enterprise participation in defense production is nothing new. The real turning point came after World War II—the U.S. government deliberately built a standing armaments industrial system. Under that model, the government funded weapons R&D almost entirely, the underlying technology belonged to the military, and the government had the right to decide to whom companies sold and for what purpose. When Lockheed or Northrop developed a fighter jet, the Pentagon was the one paying and the owner of the technology.

The logic of the new generation of Silicon Valley defense companies is entirely different. They claim to develop weapons with private capital, usually venture capital, and then sell the finished products to the government like commercial products. The founders of Palantir and Anduril repeatedly emphasize this narrative: we don't need you to tell us what to build; we fund the R&D ourselves, and you just buy it off the shelf. This model gives companies greater control over the underlying technology, but it also plants the seeds of power tension between the government and private capital.

The Starlink Incident: When a CEO's Personal Judgment Sways the Battlefield

This tension has already played out in real life on the battlefield in Ukraine. Ukrainian forces rely heavily on SpaceX's Starlink and Starshield satellite internet for communications and operations, but recent reports say the services were slowed down, materially affecting Ukrainian military actions. Musk denies that this was his decision, but the incident itself exposes a structural problem: when critical military communications infrastructure is in the hands of private companies, the boundaries of wartime decision-making become blurred.

Weinberger traces a similar incident back to 2022: Ukraine wanted to use Starlink to connect armed naval drones to attack the Russian fleet in Crimea, and Musk refused the request. The public only learned the details when biographer Walter Isaacson disclosed them. Musk first said he didn't want to escalate the war, then changed his tune and said sanctions prevented him from providing support—this kind of "deliberate ambiguity" is precisely the problem.

Weinberger uses a sharp analogy to illustrate the absurdity: can you imagine the U.S. Secretary of Defense having to call Lockheed Martin's CEO to ask permission to use the F-35 before launching a military operation? That could never happen. Traditional defense contractors have clear "guardrails"—their behavior is constrained by their relationship with their government customer, and CEOs don't casually air personal views on global security issues on social media. But tech titans like Musk, Peter Thiel, and Palmer Luckey, who have independent sources of capital, face no such constraints.

In-Q-Tel is a nonprofit venture capital firm established by the CIA in 1999, with the mission of helping the intelligence community access cutting-edge commercial technology and avoid missing out on Silicon Valley innovation due to rigid procurement processes. It does not directly inject capital into companies in exchange for equity; instead, it acts as a "strategic investor," using small amounts of money to leverage technology partnerships while serving as a bridge between the CIA and startups. Palantir's early contract from In-Q-Tel is a typical example of this model: the CIA injected seed funding into Palantir through In-Q-Tel and provided its first government customer, giving the company crucial endorsement before it had proven commercial viability. In-Q-Tel's existence shows that the U.S. intelligence apparatus tried to engage private technology using venture capital logic well before the "Silicon Valley defense revolution" narrative took shape, and its experience and lessons form important background for understanding the current defense procurement transformation.

Apple Was Once "Kept Alive" by the CIA: A Forgotten History of Silicon Valley and Defense

Weinberger unearths a little-known episode in her book that serves as a key entry point for understanding the relationship between Silicon Valley and national security agencies.

In the mid-1980s, after being ousted from Apple, Steve Jobs founded NeXT, which made expensive workstations (about $10,000 apiece). The company struggled. Jobs personally pitched DARPA, and then-director Vic Reis was persuaded by his "reality distortion field" and bought a few dozen units. But the real big customer was hidden elsewhere—the National Security Agency (NSA) reportedly had a much larger contract.

After years of digging, Weinberger found that it wasn't actually an NSA contract but a CIA contract. A former senior CIA official confirmed that they purchased tens of thousands of NeXT workstations for satellite imagery analysts, with the contract running from the late 1980s into the 1990s. Multiple sources familiar with the matter agree that this contract gave Jobs years of runway to prove himself and his technology, ultimately allowing him to return to Apple. And NeXT's operating system later became the foundation of Apple's product operating systems.

This story reveals two key points: Silicon Valley got its start on military and intelligence money; and when Jobs returned to Apple, it coincided with the end of the Cold War, defense spending cuts, and the explosion of the personal computer market, making consumer technology the bigger gold mine—and the entire Silicon Valley drifted away from the Pentagon. The so-called "Silicon Valley anti-war" cultural impression is largely just the experiential limitation of a generation of tech workers who grew up in peacetime—they didn't participate in defense because there was no market for it at the time.

After 9/11: How Palantir and SpaceX Cracked the Pentagon

After 9/11, the U.S. government realized its IT systems were outdated and disconnected. Some visionaries began to act. Peter Thiel, having profited from the sale of PayPal, applied anti-fraud algorithms to counterterrorism data analysis and founded Palantir. The company got its first contract from In-Q-Tel, the CIA's venture capital arm. Weinberger points out that Palantir's early true disruption wasn't some "magic AI" but a simple fact: it offered a point-and-click, easy-to-use interface. Government systems at the time were clunky and hard to use, and an intuitive front end was itself revolutionary.

Musk took a different path. SpaceX's core selling point to the Pentagon wasn't the most advanced or powerful rocket but lower launch costs. This was equally disruptive for a military long plagued by high launch costs and schedule delays.

Weinberger is quite direct in her criticism of the traditional defense procurement system: "It's a mess." A weapons system often takes decades from requirements definition to delivery, and cost overruns are the norm. This gave Silicon Valley newcomers ample room to disrupt.

The Defense Production Act (DPA) is a core legal lever for U.S. government intervention in private industry. Originally enacted in 1950 during the Korean War, it authorizes the president to require companies to prioritize defense orders, expand production of critical materials, and even exercise eminent domain over some production facilities when national security requires it. The law has been amended several times and remains the primary basis for federal mobilization of the industrial base. But in highly complex areas like weapons systems, the law's practical effect is limited: it can compel companies to accept orders and prioritize supply, but it cannot force them to hand over the underlying technology or intellectual property they developed independently. This is precisely the institutional backdrop for the red-line controversy involving AI companies described in the article—the government has the power to demand "what to produce," but lacks direct authority over how privately developed technology is used, and the case of Truman's steel mill takeover being struck down by the Supreme Court delineates the constitutional boundaries of presidential power.

"Fake It Till You Make It" Meets Live Fire

Anduril is a typical case for testing this new logic. Founder Palmer Luckey got his start with the virtual reality headset Oculus, and the company pivoted to defense after selling Oculus to Meta. Weinberger cautions that Luckey has yet to score a mass-market consumer success—Oculus never became a mainstream product. And Anduril started with nothing more than a PowerPoint deck made at Luckey's home, claiming it would build aircraft, missiles, and various technologies, with no actual products at the time.

The company's valuation now exceeds that of traditional defense giant Northrop Grumman, but product performance is questionable. Anduril's underwater drone was once captured by Iran, and the government's explanation was that the device malfunctioned—hardly a good advertisement for an arms maker. Altius and other drone models deployed to Ukraine have also had problems, including being jammed by Russian forces.

Weinberger points out a structural feedback dilemma: when U.S. companies' products developed at home fail in Ukraine, sending them back to the U.S. for repair takes a year, while Ukrainian teams can make fixes on site in hours to days. This gap in iteration speed poses a substantive challenge to U.S. companies, including Anduril.

Even more alarming is a certain real mindset in venture capital circles. Weinberger interviewed a prominent defense-focused venture capitalist who said bluntly: I don't care whether this thing works, as long as the government buys it. When pressed about a crewed hypersonic vehicle that experts generally consider decades away from being possible, the investor's response was: I don't invest in things that work; the Department of Defense has always bought junk, they don't care about quality, so why should I?

This is precisely the dark side of Silicon Valley's "fake it till you make it" culture in the defense realm. In consumer internet, if a product fails, users just switch apps; on the battlefield, soldiers have only one chance.

"Fake it till you make it" is a signature creed of Silicon Valley startup culture, referring to startups raising money, recruiting talent, and winning customers with grand visions before their products are mature, relying on rapid iteration to fill in capabilities along the way. This strategy is low-cost in consumer internet: if a product fails, users switch to a competitor, and the losses fall on investors and the company. But transplanting it to defense fundamentally changes the risk structure—failure of a weapons system can mean casualties or mission failure, iteration cycles are measured in years rather than weeks, and battlefield conditions cannot be A/B tested like an app. The combination of the VC mindset and this culture that the article reveals constitutes the sharpest challenge to the "Silicon Valley model": when the object of "faking" shifts from user growth curves to missiles and drones, the margin for error is compressed to nearly zero.

When AI Companies Start Drawing Red Lines for the Pentagon

The interview also touches on a brewing conflict: the contest between AI companies and the military over the boundaries of technology use. According to Weinberger, Anthropic, after a contract involving military operations, demanded two red lines be added—its large language models must not be used for mass surveillance of Americans and must not be used in fully autonomous weapons systems. The Pentagon reacted fiercely, not only terminating the contract but also listing Anthropic as a supply chain risk.

Weinberger takes a nuanced position: she understands the military's anger but also points to a fundamental reality—when companies develop technology with private capital, the government cannot force requisition the way it can with government-funded projects. During the Korean War, Truman tried to seize steel mills to stop a strike, and the Supreme Court ruled the president had no authority to do so. The government has levers like the Defense Production Act, but ultimately it cannot arbitrarily order private companies around.

The core issue isn't whether Anthropic's ethical stance deserves praise, but that: the red lines you agree with today may be replaced by red lines you disagree with tomorrow. If the person deciding to restrict Ukraine's use of a certain technology is Musk—out of his concern about escalating conflict with Russia—would you still think this mechanism is reasonable?

Supply chain risk is a formal designation with legal force in the context of U.S. defense procurement, usually meaning that a supplier is deemed unsuitable to participate in defense projects due to foreign ownership, security concerns, or unreliability, and may face contract termination, procurement exclusion, and other consequences. Labeling a domestic AI company as a supply chain risk is far more serious than an ordinary commercial dispute: it means the Pentagon is not only excluding the company from the current contract but may also affect its eligibility to participate in all future defense projects. This tool has traditionally been used against foreign adversaries or suppliers with espionage risks; now it is being used against an American company that fell out with the military over ethical clauses, marking a shift in the defense procurement system's definition of "reliability" from technical security to political and ethical alignment—a precedent worth watching.

The Most Likely Outcome: A New Set of Names, the Same Rules of the Game

Weinberger's judgment about the future is not optimistic, but it isn't the darkest version either. She believes the most likely scenario is that Silicon Valley defense upstarts eventually become the new establishment, sitting alongside Lockheed, RTX, and Northrop. The defense industry remains a closed system, just going from five players to seven or eight.

The difference is that the new players wield far more political power than the old contractors. Traditional defense companies depend on government contracts and are constrained in their behavior; the founders of Palantir and Anduril, by contrast, have independent capital, huge retail investor bases, and the ability to speak directly to the public. Weinberger notes that Palantir's retail investor share is far higher than institutional investors, and these investors care more about the "saving Western civilization" narrative than about enterprise software fundamentals and contract details—one reason its valuation has been questioned.

What she fears is "the worst of both worlds": neither reaping the real benefits of innovation nor preventing unconstrained tech giants from gaining the power to sway decisions of war and peace.

The solution is not for the government to take over everything. Weinberger jokes that journalists are good at pointing out problems rather than providing answers, but she offers two directions: first, competition—avoid any single company forming a de facto monopoly; she notes that Musk has been described by The Wall Street Journal as a "de facto monopoly" in space launch, a situation she finds troubling; second, avoid letting tech giants take government positions meant to regulate them. She gives the example of the Pentagon appointing Musk and Luckey to a committee studying "future warfare" when both are among the largest defense contractors—an irreconcilable conflict of interest.

The interview ends on a note of optimism: in recent years, a rare consensus has emerged between left and right on the issue of limiting the power of big tech. In a deeply polarized era, this may be a faint hope for change.

One-line Summary

Venture-backed Silicon Valley defense upstarts are reshaping the arms industry—while transferring war-making decisions to unconstrained private titans.

Paragraph Summary

Based on an interview with investigative journalist Sharon Weinberger, this article traces the history of Silicon Valley's relationship with the Pentagon and reveals the structural contradictions behind the "defense tech revolution." A new generation of defense companies develops weapons with private venture capital and sells them to the government off the shelf, changing the post-WWII procurement model in which the government fully funded and owned the underlying technology, giving companies greater control over military technology. Events on the Ukraine battlefield—Starlink service being slowed and Musk refusing to authorize an attack on Crimea—expose the risk that private CEOs can materially influence the course of war. Anduril's valuation exceeds Northrop Grumman's, yet its products' combat performance is questionable, and VC circles harbor a "as long as the government buys it" mentality. AI company Anthropic was listed as a supply chain risk after demanding red lines on surveillance and autonomous weapons in its contract, highlighting that the government cannot force the requisition of privately developed technology. The author judges the most likely outcome to be Silicon Valley upstarts becoming the new establishment, producing "the worst of both worlds."