How Silicon Valley’s ‘Oppenheimer’ found lucrative trade in AI weapons

From below, the ALTIUS-600M looks like a black cross sailing through the sky. It has a slender fuselage and a stretched wingspan that makes it twice as wide as it is long. Up close, it’s a blend of metal and circuitry, a sleek and ominous testament to human ambition. It is also one of the first artificially intelligent weapons to be deployed by the US in a real war. Shot from a tube like a missile, first its wings telescope outwards, then it identifies its target, flying for up to 280 miles. It circles high in the sky, for as long as four hours, and then strikes on the ground. The “M” is for munitions; on impact it explodes in a ball of flames.

The attack drone is made by Anduril Industries, a defence company founded by a 31-year-old Californian inventor called Palmer Luckey. The US government has bought hundreds of ALTIUS-600Ms as part of its military aid packages to Ukraine. It is one of more than a dozen autonomous defence and weapons systems Anduril has built and sold to the Pentagon since it was founded in 2017. In that time, Luckey’s company has become one of the largest of only a handful of start-ups that have infiltrated the US government’s trillion-dollar war machine. The industry has been dominated for decades by an oligopoly of century-old contractors, known as “primes”, such as Lockheed Martin and Raytheon.

As conflicts multiply overseas, Anduril’s business is accelerating: annual revenues are projected to reach $1bn by 2026. “The first page of our first pitch deck said that Anduril will save western civilisation and save taxpayers hundreds of billions of dollars a year as we make tens and tens of billions of dollars a year,” Luckey says. “The intent is to go toe to toe with the major primes and try and fight our way to an equal footing.”

It is a braggadocious claim delivered with the conviction of someone who has predicted the future before. Anduril is Luckey’s second big venture. In 2014, when he was 21, he sold Oculus, the virtual reality headset business he started in his parents’ garage, to Facebook for $2bn. The next year, he was on the cover of Time magazine. Oculus was a technological turning point, the first time anyone had created a viable consumer VR product. It changed the trajectory of Facebook, one of the world’s largest companies, as it reoriented towards the “metaverse”. And Luckey’s invention was the foundational model that spurred headsets like Apple’s recently released Vision Pro. He represented a classic Silicon Valley fantasy: the dishevelled nerd destined for greatness.

Anduril’s ‘ALTIUS-600M’ attack drone © Maggie Shannon

But by the time Luckey started Anduril — named after the “Flame of the West” sword in JRR Tolkien’s The Lord of the Rings — he had gone from wunderkind to pariah. After revelations that he donated to an anti-Hillary Clinton group online, he was fired from Facebook, and his support for Donald Trump alienated him from a liberal tech industry still reeling from the outcome of the 2016 presidential election. The money to incubate Anduril came from Peter Thiel, the controversial co-founder of PayPal, who had also been disavowed by parts of Silicon Valley for his donations to the Trump campaign and for driving Gawker Media into bankruptcy. “I’m a big Peter fan,” says Luckey. “He understood what it was like to get into business trouble for political reasons.”

Anduril’s first big win, when it was just a few months old, was a contract with the Trump administration to supply AI-powered surveillance towers along the Mexican border. It was a triple whammy of anathema to the high-tech elite, who were scrambling to define their ethical position on AI and retreating from selling software to the military to stave off employee protests. And yet, seven years later, Luckey has arguably become the most crucial figure bringing Silicon Valley to the front lines of American national security.

Wars in Ukraine and the Middle East and a growing threat from China and Russia have accelerated urgently needed modernisation of military capability in the west. In 2021, China started testing hypersonic weapons — projectiles capable of travelling five times the speed of sound — a significant threat demonstrating technological advances that have left the US far behind. This month, Vladimir Putin warned that western support of Ukraine risks triggering a global nuclear war.

So Silicon Valley has been dragged back to the heart of a military industrial complex that has lost its edge — and Luckey is positioning Anduril as its saviour. Advanced weapons systems that sounded like science fiction a few years ago are now being built at Anduril’s headquarters in the suburbs of Los Angeles, and a new generation of defence technology companies is emerging in its orbit. At the top of Anduril’s mission statement, Luckey poses the question: “Xi Jinping thinks he can out-innovate American defence. Is he right?”

On the February day I meet Palmer Luckey at Anduril’s sprawling campus in Costa Mesa, Orange County, there has been an escalation in the Middle East. Overnight, three American soldiers have been killed in Jordan by an Iranian drone, the latest sign that the Israel-Hamas war threatens to boil over into a larger conflict. Details are scant, but early reports suggest the drone was able to drop into the US base undetected by counter-drone technology.

Luckey, in his trademark Hawaiian shirt, shorts and flip flops, runs his hands through his mullet in exasperation. Anduril’s technology is more advanced, faster and cheaper to make, he says, than the stuff built by the leading suppliers to the US armed forces. Its biggest competitor in the counter-drone space is defence prime Raytheon’s “Coyote”. Luckey named Anduril’s system “RoadRunner” after the classic Looney Tunes bit in which the bird is always quicker than the coyote trying to catch it. “What if we had counter-drone technology deployed to all our US footprints that made that attack totally futile?” he says. “You could send a hundred times more drones and they would just get swatted out of the sky, and your adversary would spend tens or hundreds of millions of dollars attacking the site for absolutely no tactical gain.”

A grandiose and unwavering belief in the power of deterrence infuses Anduril’s mission. Essentially, Luckey’s aim is to make the US and its allies almost impossible to harm — “a prickly porcupine” in his words — as well as to supply weapons powerful enough to put adversaries off attacking in the first place. “We want to build the capabilities that give us the ability to swiftly win any war we are forced to enter,” he says.

The thesis is not original. It’s the same idea that led to Robert Oppenheimer’s development of the atomic bomb. America’s attacks on Hiroshima and Nagasaki ended the second world war, killing more than 100,000 people, and nuclear weapons have not been deployed in conflict since. The problem is that, 80 years later, to truly determine the outcome of a war, the US would not only need to have war-ending technology but also convince its adversaries it is willing to use it.

That is where artificial intelligence, often referred to in defence circles by the less ominous-sounding “autonomy”, comes in. “The thing that’s so powerful about autonomy is that you can clearly show your adversaries that you have weapons that do not cost all that much money and that don’t cost human life,” Luckey says. “It’s a powerful part of deterrence that the US has lost over time as our willingness for death has gone down and the cost of our systems has gone up.”

Anduril CEO Palmer Luckey beside his ‘RoadRunner’ drone
Anduril CEO Palmer Luckey beside his ‘RoadRunner’ drone © Maggie Shannon

All of Anduril’s tech is capable of operating almost entirely unmanned (discretion as to how autonomous it is when deployed is down to users and policymakers). The products all run on the same machine learning software — an operating system called “Lattice” — that means they can communicate with each other and be updated like any smartphone to keep pace with advances in enemy technology and intelligence. This is part of the reason Luckey says his systems are better than, say, Raytheon’s.

The sales pitch is clear: commercial technology develops at breakneck speed amid the competition of the free market; defence doesn’t. There is better AI in a Tesla than in US military vehicles, Anduril claims, and far greater computing power in an iPhone than the systems regularly used by the Department of Defense. Until 2019, America’s nuclear arsenal still operated off floppy disks. Compared to that, Anduril’s autonomous fighter jet, Fury, and its “undersea battlespace” submarine, Dive-LD, seem as improbable as something lifted from the pages of a Jules Verne novel.

Now that pitch has a fresh sense of urgency, says Luckey: “I’ve been joking that the past few years have been the ‘Palmer Luckey I told you so’ tour. This was the thesis of the whole company . . . Our first pitch deck to investors specifically included what Russia was going to be doing, what China was going to be doing . . . how Iran would use proxies to try to suck us into a fight. It’s exactly as we predicted.”

On a summer evening in San Francisco in 2019, General Catalyst’s 40 partners were summoned to a meeting at the venture capital firm’s downtown office. Anduril was doing its first big fundraising, did General Catalyst want in? The atmosphere was tense. The partners had cut a tiny, $1.5mn cheque to the fledgling company two years before, but now they were divided over the ethics of giving tens of millions of dollars to what was by then an all-out arms manufacturer. “It was contentious,” says one of the partners, who didn’t want to be named. “Palmer was, to put it mildly, not a popular figure at the time.”

Silicon Valley’s early history might have been lost on those naysayers. The legacy of the weapons developed to help win the second world war — the atomic bomb, jet power, radar — was a profound awareness that science and technology would determine dominance on future battlefields.

Anduril’s autonomous kinetic interceptor
Anduril’s autonomous kinetic interceptor © Maggie Shannon

As the US raced for ideological and military superiority over the Soviet Union, it poured billions of dollars into research. In 1948, Project Rand (a contraction of “research and development”) was born to connect military planning with American scientific discovery. Rand’s ranks of consultants and big thinkers produced the building blocks of everything from the personal computer to today’s AI revolution.

A decade later, the Soviet Union’s launch of the world’s first satellite, Sputnik, sent shockwaves through America. In 1958, President Dwight Eisenhower set up Arpa, the Advanced Research Projects Agency, to provide money to develop technologies for military use and make sure the US would never be out-innovated again. Darpa (the “D” for defence was added later) funded 70 per cent of all US computer research in the early 1960s, much of it coming from labs in the Californian hills around Stanford University. That research led to the invention of most of the tech we use today — the internet, touch screens, GPS, self-driving cars — and ushered in a new era of American technological supremacy.

When foreign leaders have, over the years, vowed to foster their own Silicon Valleys, they were not merely talking about consumer-facing tech companies like Twitter and Airbnb. What they really want to replicate is the American framework of government and private companies working in concert to maintain national economic and military dominance. Even so, it is hard to overstate how unlikely a proposition a Silicon Valley defence company would have sounded to many tech investors even five years ago.

Silicon Valley began drifting away from its roots in the late-1990s and 2000s, with tech companies adopting mottos such as “Don’t Be Evil” and telling investors, employees and customers they existed to make the world a better place. By 2019, memories were still fresh of Google and Microsoft’s embarrassing retreat from some military work, when employees staged walkouts over contracts with the Pentagon.

Three years into the Trump presidency, the tech industry was suffering from near-constant reactionary spasms to Washington’s policies, its progressive ideals clashing with Trump’s executive order on Muslim immigration, his equivocation on white supremacy, his ban on transgender troops. In return, DC was picking fights with Big Tech, accusing companies of stifling conservative viewpoints on their sites and hauling executives like Mark Zuckerberg and Jack Dorsey into Congress to defend themselves.

All of which meant almost no one in the Valley was doing anything like Anduril. Elon Musk’s rocket-building, satellite-launching SpaceX and Peter Thiel’s data intelligence business, Palantir Technologies, were making a killing on government contracts, including with the defence department. Anduril wanted to go a step further by building the hardware for waging war, including weapons. Its founders were warning about future conflicts with China, Iran, Russia.

Luckey was the boy-genius who had invented the Oculus headset, and Anduril’s founding team included several of the people who had helped take Palantir — also named after a piece of kit from The Lord of the Rings — from a California start-up to its then $20bn valuation. These included its chairman, Trae Stephens, chief executive Brian Schimpf and chief operating officer Matt Grimm.

One of Anduril’s early champions was Katherine Boyle, a former Florida pageant queen and ex-Washington Post reporter who became a rising star at General Catalyst. In 2017, she’d introduced her bosses to Stephens and persuaded them to write Anduril a cheque — a coup for someone so junior. Then suddenly she found she was a social pariah: “I was called a warmonger and worse,” she says. “There were times when I would get yelled at at venture capital cocktail parties.” It was not a happy time for defence tech. Anduril decided to locate its headquarters hundreds of miles south of Silicon Valley in 2017; in 2020 Palantir packed up its Palo Alto offices and relocated to Denver, Colorado.

Yet there were already signs that a shift was taking place in Silicon Valley. For some, the Google and Microsoft protests felt like tech company culture was under siege by activist employees. Provocateurs like Musk were garnering more attention. Big tech companies including Facebook, which had previously encouraged its employees to speak out on issues involving the company, started restricting spaces for political discussions. Increasingly, important figures in the tech industry were adopting a more conservative position. Among them was Marc Andreessen.

As tech veterans go, Andreessen is a beast. In the 1990s, he created the most popular web browser of the early internet age, Netscape Navigator. Now he runs a venture capital firm with his partner, Ben Horowitz. He is a commanding, often belligerent, evangelist for technology, whose confrontational exchanges on social media and outspoken criticism of government and taxes have made him controversial. But his influence in Big Tech, including at Facebook where he sits on the board, has kept him in a power position in Silicon Valley. When he speaks, people tend to listen.

Luckey at Anduril’s headquarters in Costa Mesa
Palmer Luckey at Anduril’s headquarters in Costa Mesa © Maggie Shannon

After Andreessen Horowitz invested in Anduril in 2019, the mood changed fast. When he wrote on his website, “I believe in the United States of America. I believe in a strong national defence. And I believe in Anduril,” it was the first full-throated statement from an authoritative insider that investing in defence technology was permissible. It was a game-changer for VCs who had been prevaricating on the sidelines, wondering how they would justify defence investing to backers and staff. “Silicon Valley has a huge culture of following the latest trend,” says Luckey. “All it takes is a few people to stand up and say, ‘Attention everyone, the acceptable set of beliefs has changed,’ for things to change for everybody. The pendulum swings back and forth, and I think there’s a lot of people who think the pendulum swung too far in one direction on a whole variety of social and political issues.”

Since then, Anduril has raised about $2.8bn. Thiel’s venture firm, Founders Fund, put in $400mn, its single largest cash investment into any company. General Catalyst eventually decided to invest, too, handing over $30mn in 2019. Since then, there has been a gold rush for start-ups making defence technologies; VC investment into defence tech doubled to $33bn between 2019 and 2023. Confidence is growing that defence tech start-ups will eventually take a significant share of the US’s defence budget, which stands at a hefty $842bn this year.

Planning for your start-up to make tech for military use is now firmly in vogue in Silicon Valley. Earlier this year, OpenAI — the maker of generative AI bot ChatGPT — quietly changed its terms of service to allow military applications of its technology, deleting its previous policy that prohibited use for the purposes of “military and warfare”.

Luckey’s politics remain unchanged. He self-describes as a “big R” Republican and a “little L” libertarian. He’s non-interventionist, but this is not as counterintuitive as it sounds. He is developing autonomous military systems to keep America from being caught up in foreign wars, he maintains. “I don’t want boots on the ground. I’m tired of the US getting dragged into fights that our allies really should have the tools to fight with themselves.” Luckey implies he will vote for Trump again this year, noting that he is a longtime supporter. In 2011, he wrote to the then reality-TV star asking him to run for president. In 2020, he hosted a fundraiser for Trump at his home, which the then president attended. “I talk with him from time to time,” Luckey says of Trump. “I like to think he thinks I’m a competent guy, but I don’t think I’m Trump’s defence whisperer. That would be fantastic if it was true, I’d love to have that level of influence in US policy.”

Seto Kaiba is the anti-hero of the Japanese manga series Yu-Gi-Oh! — an intelligent and Machiavellian orphan who inherits a weapons empire and pivots it into virtual reality technology. He is also Luckey’s personal hero. When he was seven, Luckey became obsessed with one of the character’s lines: “You said tech has limits. Wrong.” (As Luckey puts it, Seto Kaiba then proceeds to “kick everyone’s ass using his incredible technology”.) “Sometimes I wonder, how much free will do I actually have over my life?” Luckey tells me, after animatedly recounting this scene. “Is it possible that I actually had no choice at any point but to pursue VR and weapons development?”

Luckey’s interests have guided him down a highly specific path. His grandfather, a pilot who was interested in military aviation, started passing down history books when Luckey was eight or nine. Then he got into science fiction. “Most of the stuff Anduril is working on is stuff that has existed for decades in the realm of science fiction,” he says. He cites Verne, who predicted plausible future technologies like electric submarines and the Taser when he was writing in the mid-1800s.

Homeschooled by his mother when he was a teenager, Luckey also devoured video games and manga. He grew up in Long Beach, California, with his father, a car salesman, and three sisters, one of whom, Ginger Luckey, is married to Florida Republican congressman and Maga diehard Matt Gaetz. By the time he was 11, Luckey was experimenting with electronics, first in his parents’ garage, then in an old trailer in the driveway, fashioning lasers and an electromagnetic coil gun from copper wire and high-voltage capacitors.

Anduril’s autonomous underwater vehicle
Anduril’s autonomous underwater vehicle © Maggie Shannon

Once, when he was working on a Tesla coil, he touched a grounded metal bed frame and blew himself across the garage. He also burnt a grey spot into his vision while he was cleaning an infrared laser. Later, he started modifying gaming equipment and fixing broken iPhones, tinkering with old military VR headsets that he bought through government auctions. Diving deeper, he got into the habit of reading declassified Pentagon research papers. First anything ever written on head-mounted displays and virtual reality, then vast quantities of technical reports from government-funded labs, going back to the second world war. They are the “most fascinating reads you could find because it was a time in history where you could think without limitation or regard for regulation or precedent”, he says. “They were living in the atomic age, this new Americana-driven world. Anything was on the table.”

This hyper-techno-optimism consumes Luckey’s personal life too. He drives a Tesla model S (with the paint stripped off to reveal the bare metal) and sails an 82ft Mark V special operations craft, a $3mn camouflaged Navy Seal vessel mounted with decommissioned machine guns, which he keeps docked at his Newport Beach, California, home. In 2022, as a side project to commemorate the anime series Sword Art Online, he invented a VR gaming headset that actually kills the player if they die in the game, by mounting explosive charge modules to the headset aimed at their brain. (He says he has never put it on.)

Luckey and his wife Nicole Edelmann, a professional gamer he met at debate camp when they were teenagers, have been photographed wearing bikinis and combat boots, both cosplaying as Quiet, the heroine of Metal Gear Solid. In that video game, players uncover the existence of an AI-controlled nuclear weapon whose villainous creator built it out of the belief that human unwillingness to launch a nuke is the flaw in deterrence theory.

“I genuinely think this guy is a genius,” says Chris Brose, Anduril’s chief strategy officer, who was Senator John McCain’s principal adviser on national security, and is the company’s most senior hire from Washington DC. “He comes up with a solution to a problem and understands immediately technically how to do it — and 80 or 90 per cent of the time he is right, right out of the gate.” When they were designing Anduril’s first counter-drone interceptor system, Anvil, Luckey envisioned a quadcopter that could knock other quadcopters out of the sky, says Brose. “He had a prototype in a week . . . He has ideas constantly. Some of those ideas are going to be wrong or unworkable. Others are going to be Oculus.”

A few weeks after Russia invaded Ukraine in February 2022, Luckey flew to Kyiv and met with Volodymyr Zelenskyy, for the second time. They first talked in New York three years earlier, after the Ukrainian president read an article in Wired magazine about Anduril’s sensor towers on the US-Mexico border. Zelenskyy had wanted to deploy the same tech on Ukraine’s eastern border, according to Luckey. “People will try to paint him as just a comedian,” he says, “but the reality is years before Russia invaded it was one of his top priorities to the point where he was visiting foreign companies to figure out how to stop it.”

The US government blocked Anduril from selling the technology to Ukraine. The official American position was that Russia was merely sabre-rattling and that a build-up of technology on the border, particularly supplied from the US, would be provocative. “I actually believed [Zelenskyy’s] assessment was right. But we have to play by the rules, so we weren’t able to do anything about that,” Luckey says. When the two met again in Kyiv at the start of the war “there was a little bit of sourness” from Zelenskyy about that decision. (Zelenskyy did not respond to requests for comment.)

The Ukraine war has been a useful test-ground for Anduril to hone its systems in a real-life scenario. More importantly, the conflict shows how the characteristics of war are changing. More commercial technology, like satellite imagery and autonomous drones, is being used than ever before. When Musk deployed his Starlink service to Ukraine, providing internet despite Russian jamming efforts, it was the first time a commercial company had provided the backbone for a country’s military capability during wartime. That gave weight to an argument that has been brewing in DC for years that the US needs to get better at buying technology from outside traditional procurement. The sheer quantity of munitions needed for the fighting happening in Ukraine — which is currently going through 10,000 drones a month — exposed the vulnerabilities with US defence production. The need for the US to acquire new technologies quickly and at scale turned from an academic point to consensus overnight.

SpaceX and Palantir have both complained about how contracts are awarded going back as far as 2005, including suing the US Air Force and the US Army respectively over processes they claimed unfairly favoured legacy providers over newcomers. After the collapse of the Soviet Union, the US slashed defence spending, kicking off a frenzy of consolidation in the early 1990s, when dozens of companies folded into just five. Now, as defence spending surges, about 80 per cent of the industry’s revenue is produced by just 10 companies, founded decades ago. Nearly two-thirds of major weapons systems in the US have a single bidder.

At the heart of the SpaceX and Palantir lawsuits was consternation over the defence department’s buying model known as “cost-plus”, in which it awards contracts to companies that offer to build at the cost of production plus a percentage fee. This typically leads to big cost-overruns and long delays as contractors are incentivised to allow costs to stack up. It also means that systems are refreshed on a generational basis, far too slowly to keep up with technological innovation. For example, the Lockheed Martin F-22 Raptor introduced in 2005 was the first jet fighter to incorporate advancements in digital computation and mobile networking in the 1990s. The next generation of the jet is not expected until the 2030s.

Unlike traditional prime contractors, SpaceX, Palantir and Anduril use private funding to build their products before selling them to the government. As each company began to win billions of dollars’ worth of contracts, it’s started to look like the defence department’s buying model is slowly becoming more commercial. But a level playing field is still far off.

There are certain parts of Anduril’s headquarters where Luckey has to wear proper shoes. In 2021, Anduril moved into the former Los Angeles Times printing press, taking over the nearly 650,000 sq ft site, about the size of 11 American football fields. Its roughly $200mn redevelopment has created a vast engine of research, development and manufacturing. It has the same hum that imbued the postwar research labs of the San Francisco Bay Area.

From the outside, the buildings are slate grey and discreet. Security guards are posted at each of its street entrances (my taxi driver guesses the company has “something to do with the government”). Anduril has a dotcom-era bring-your-dog-to-work policy, and employees walk their pets around the campus in the southern Californian sun. Inside, the latest experiments are being worked on in aircraft hangar-sized warehouses — a blast-proof, satellite-clad shipping container and a desert-camouflage all-terrain vehicle, covered in sensors. Black screens on wheels shield secret projects at one end of the room. The dogs are in here too, dozing under workbenches and peering out of meeting rooms. A sign with crossed-out flip flops is aimed squarely at Luckey. On the manufacturing floor, only real shoes will do.

Palmer Luckey in the boardroom at Anduril’s headquarters in Costa Mesa
Palmer Luckey in the Anduril boardroom © Maggie Shannon

When Facebook and Google built their palatial office compounds in Silicon Valley, they were multibillion-dollar symbols of Big Tech’s immense power and status. Anduril’s campus is equally a testament to the shift in attitudes in the Valley since then. Whereas a decade ago, Anduril would have found it impossible to recruit from the same pool of developers and engineers that were destined for commercial tech companies, now it is pulling employees from those rivals. Anduril’s staff has tripled in size to more than 2,500 in three years.

New recruits arrive with zero ambiguity in their minds about the technology they will be working on. Every new hire sits through a three-hour presentation by Luckey about Anduril’s story and its plans for the future. The message is: if you are not comfortable, you should not work here. There’s something else here too, a different answer to the question of what Silicon Valley is for. The golden age of Netflix and Apple brought life-changing consumer technologies. But by the 18th version of the iPhone, the gains seem marginal compared to marketing copy promising magic. Even the new Apple headset, which is light years ahead of anything Luckey was tinkering with in his garage, seems, well, kind of mid. If Silicon Valley is a country of its own, it seems to be one that is running out of significant ideas. But if, as it seems on Anduril’s campus, Silicon Valley is an asset of American power, then Luckey will have been right a third time.

His ideas are only going to get more ambitious as AI unlocks new generations of weapons systems, and as Anduril raises billions of dollars more capital. “If you are so incredibly strong that you can fight with one hand tied behind your back and still easily win, you can actually afford to say, ‘I’m not going to accept civilian collateral damage to nearly as great of an extent, I’m not going to accept massive damage to the US economic machine,’” Luckey says. “The way to frame it would be that I want to give ourselves a technology that turns the world stage, as it pertains to warfare, into the United States being an adult in a room full of toddler-sized dictators.”

In his mind, Luckey already is Seto Kaiba, capable of kicking everyone’s ass with his incredible technology. So will Anduril create the AI weaponry that will be as decisive to the outcome of war as the atomic bomb? “We have ideas for what they are,” Luckey says. “We are working on them.”

Tabby Kinder is the FT’s West Coast financial editor

Follow @FTMag to find out about our latest stories first and subscribe to our podcast Life and Art wherever you listen

Video: Can generative AI live up to the hype? | FT Tech


Leave a Comment