“I don’t know what you want to call the conflict we’re in with China and Russia, but I’d say we’re at war.”
Michael Brown, director of the Defense Innovation Unit (DIU) within the U.S. Department of Defense, declares war rather matter-of-factly. “It’s just not war in the way we typically think,” says the longtime Silicon Valley executive who transitioned into government service in 2016. But when adversaries get to the point of “shutting down pipelines and creating other damage,” he says, you’re seeing the “kinetic effects of cyber.”
When Brown and I chat in August—at a picnic table in a glade of cedar trees behind a tan brick building in Mountain View, California, that houses both the 341st Readiness Division of the U.S. Army and the DIU—news continues to dribble out about the extent of the 2020 SolarWinds hack. The event involved a group reportedly backed by the Russian government that compromised networks of a number of U.S. government agencies, including the Department of Energy and the National Nuclear Security Administration. A couple of weeks earlier, Secretary of State Antony Blinken, along with a coalition of U.S. allies, accused China of engaging in cyber espionage, and the U.S. Department of Justice charged four Chinese nationals with being part of a global hacking campaign targeting companies, government agencies, and universities. Chinese officials vehemently denied the charges.
“For cyber, for electronic warfare, for misinformation,” Brown says, “we need to be shifting more dollars to those domains.”
Brown’s job is to try to make that happen. As head of the DIU, he runs what’s designed to be an agile skunk works within the Department of Defense (DOD). It identifies commercial tech products that it believes the DOD needs to prepare for technology-driven warfare, and seeds the companies that make them with research funds. The startups that meet the government’s standards win a contract with a service branch such as the Air Force. Some of the companies the DIU has worked with include Anduril Industries, the drone and security startup founded by Oculus VR creator Palmer Luckey; C3.ai, enterprise-grade predictive tech founded by longtime entrepreneur Tom Siebel; and even Oura Health, maker of a ring that tracks a wearer’s sleep and activity, to anticipate soldiers’ illnesses before they show symptoms.
The DIU was created in 2015, during the Obama administration. Brown, who arrived in Silicon Valley four decades ago as a student at Stanford’s graduate school of business and rose to become CEO of the security company Symantec, has run it since 2018. His success led President Joe Biden to nominate him in April to be the DOD’s undersecretary for acquisition and sustainment, the Pentagon’s top weapons buyer. In July, Brown withdrew his nomination amid an ongoing investigation involving a claim that he showed undue favoritism in his hiring practices. He denies the charge and declines to discuss it.
The DIU’s success has spawned scores of other such innovation cells within the Pentagon. Although they operate independently of one another, together they reflect a distributed effort to modernize national defense—and a realization that the U.S. government needs to make history rhyme, echoing the efforts that ultimately won the Cold War with the Soviets by relying on the microwave and chip manufacturers clustered in the flatland south of San Francisco. At that time, in the 1960s, the DOD was funding twice as much R&D as U.S. businesses were, and it enlisted what came to be known as Silicon Valley to design and manufacture the processors that would guide missiles, launch satellites, and take humans to the moon. Lockheed (now Lockheed Martin) set up its division to build Polaris missiles in Sunnyvale, and Fairchild Semiconductor, the company that in many ways spawned the modern technology industry, made 80% of its early revenue from the military. “We were the investor, we were the first customer, we were an early adopter,” Brown says.
But then, starting in the late 1970s, regulatory changes enabled the rise of venture capitalists, who pushed tech startups to build businesses at a pace incompatible with the agonizing, red-tape-laden process required to win and fulfill government contracts. In the Reagan era, the military increasingly tailored national defense around massive platforms such as aircraft carriers that could be built only by the largest defense contractors, known as primes, such as Northrop Grumman and Raytheon.
Today, though, the technology that will likely decide 21st-century warfare will be based on artificial intelligence, autonomy, quantum computing, space, cybersecurity, and biotech. In other words, sectors in which Silicon Valley has already invested heavily to serve businesses and consumers. These technologies are being developed “at a much faster pace and with a much faster delivery than anything the primes or DOD is working on,” says Steve Blank, an entrepreneur and Stanford professor who authored The Secret History of Silicon Valley, a series about the military ties that created the region. “The Defense Department is realizing that we’ve got to have access to leading technology,” Brown tells me.
Remaking the Pentagon in Silicon Valley’s image will be far more difficult a public-private challenge than, say, Uber and Lyft steamrolling municipal taxi commissions. “You have to think about the military as a large bureaucracy that has existing relationships with primes, and they all live in a symbiosis that makes sense to them,” says Eric Schmidt, the former chairman and CEO of Google who has backed several defense startups and chaired the National Security Commission on Artificial Intelligence. Those major contractors have entrenched themselves by influencing contracting rules, spending heavily on lobbying, and sprinkling stable jobs across congressional districts.
Should Silicon Valley unseat legacy defense companies, this transformation will come with its own risks. Autonomous weapons remove military personnel from harm, but they also inoculate human operators from the suffering they inflict. Even AI’s proponents express concern that it could advance at a pace leading to a fully automated war taking place between algorithms. Two generations ago, the microchips developed in the Valley were used in the nuclear missiles that very nearly ended life as we knew it.
In exchange for that military buildup, society received middle-class jobs and eventually the commercial internet. Today, a software-focused defense industry may not yield the same civic benefits. General Dynamics, for example, has 84,000 full-time U.S. employees, located in all 50 states. Anduril has approximately 700 employees, mostly in Irvine, California.
Silicon Valley startups have targeted, and engulfed, several large economic sectors over the past decade, from education to finance. Defense is the final frontier. “We’ve affected probably 1% to 2% of DOD procurement at this point,” Brown says. “Dollarwise, I’m not expecting it to be more than 50 [percent], but it sure as heck needs to be more than 1.”
“The Silicon Valley model for building product, the iteration speed, how you invest, the economics, has become completely dominant.”
Brian Schimpf, Anduril’s CEO and a cofounder, has helped bring that pace to defense tech. The 4-year-old startup—which is focused solely on defense, has raised almost $700 million in venture capital, and is valued at $4.6 billion—is arguably the star of the DIU’s effort to bring Silicon Valley’s speed and engineering talent to the Pentagon. In July, Anduril won its 24th definitive contract, according to the federal procurement database, a five-year deal worth up to $99 million to protect U.S. bases facing drone swarms. This threat has traditionally proved hard to fend off—fighter planes are too high up to address havoc taking place 30 feet above the ground; U.S. troops in Iraq and Syria have been attacked by military-grade and, increasingly, consumer drones fitted with explosives.
In response, Anduril developed a system that keeps watch via sensors on tall poles and drones that police the air autonomously. Computer vision classifies any objects it sees, sending data back to human operators who decide what to do. Some bases dispatch humans to investigate, but Anduril also includes a specialized drone, known as the Anvil, capable of attacking an enemy drone, bodily—knocking it out of the sky. With DIU support, Anduril’s counter-unmanned aircraft system went from prototype to a product that met defense requirements in 18 months. “In a traditional DOD process,” Schimpf says, “you’d have a two-to-three-year cycle of just identifying a threat.”
The DIU has shown that it can move even faster. It worked with C3.ai, which serves both large businesses and governments with predictive analytics, to develop AI in just six months that can anticipate when Air Force planes are going to need maintenance. C3.ai’s model considers telemetry data such as vibration and rotational velocity, as well as weather information. It then builds “machine-learning models that will predict 50 or 100 flight hours in advance that this system is going to fail,” says founder and CEO Tom Siebel, who took C3 public in December 2020. “We were able to increase availability of aircraft, like, 50%.”
According to the DIU’s 2020 annual report, the group has made 208 awards since 2016, totaling almost $650 million. Twenty-six of them have resulted in a subsequent contract within the Pentagon. The DIU concept has spread across the defense agencies, each with its own tech focus, investment strategy, and a name that’s either a dizzying acronym or signifies innovation via a suffix, such as Lab or X.
One of the more notable examples, Afwerx, extends grant money to tech startups and helps match them with potential buyers within the Air Force. When Will Roper took over the program in 2018, he found ways to operate it like a VC firm, doling out $50,000 in seed funding for startups to prove their tech could work for the government, and then providing contracts up to $26 million to the ones ready to serve. Roper, who left in January and in July became CEO of Volansi, a startup that makes a vertical takeoff and landing cargo drone, made 2,300 deals during his tenure.
For all this ferment, Roper—and his colleagues—lives in fear that it’s not enough. ”[When] a new person comes along with a new idea, it’s very hard for them to get the money, the authorizations in a timely fashion before they run out of cash,” says Eric Schmidt. “This is known as the Valley of Death.”
“You know, we’re doing it out of greed and avarice, in which lies the hope of progress.”
Josh Wolfe doesn’t look like a venture capitalist, much less one who’s arguably the most active investor in defense tech. The self-described “dark cynic” dresses the part, wearing all black. “When I’m in the Pentagon wearing Vans and I put a sweater shawl over my black T-shirt, it’s a sign of signaling authenticity,” he says. “You’re getting what you see.” When we meet in September in his New York offices, he’s sporting a couple days’ scruff, and his salt-and-pepper hair crests like a wave. Fellow investor and popular podcaster Patrick O’Shaughnessy has called him the real Tony Stark. “Definitely not,” Wolfe demurs.
While many of Wolfe’s portfolio companies have won funding from within the Pentagon—he’s an Anduril investor as well as a backer of several space startups, including Orbital Insight, Planet, and Relativity Space—he is as impatient as anyone within the tech world with the Defense Department’s efforts to reinvent itself.
“It’s great that the DOD does all these [innovation grants]. You total all that money?” he muses. “It’s $1.8 billion. Look at the DOD’s budget [for R&D]. It’s $114 billion-ish. [And if] you look at what venture capital has put in technology in the last year, it’s $135 billion. Only a portion of that is directly applicable to defense, but the amount of money in sheer commercial interests to fund these things . . .”
Wolfe’s Lux Capital seeks to identify technology that could be valuable for national defense even before the DOD can. In 2016, when the firm first invested in Saildrone, which makes autonomous seafaring vessels, the startup had mostly been focused on exploration. “The original investment hypothesis was that this is going to be big for commercial, research, academic, and climate purposes,” he says. Upon seeing the high-quality observational work that Saildrone was doing for the National Oceanic and Atmospheric Administration, however, Wolfe and his partners made the connection that defense agencies would ultimately need something similar, envisioning the value of Saildrones surveying international waters near Russia, the Arctic, China, and the South China Sea. “There are strategic capabilities that [Saildrone] may have that, frankly, the Navy doesn’t.”
He’s also unafraid to go anywhere in the world to find, as he says, “what sucks, because everything around us was invented by somebody who basically said that.” A few years ago, at the invitation of General Tony Thomas, who was then in charge of U.S. Special Operations Command, Wolfe spent two weeks in Southeast Asia assessing everything from “satellite communications to blue-dot tracking of individuals.”
The trip, combined with his burgeoning friendship with “T2,” as Wolfe calls Thomas, led to Thomas joining Lux as a venture partner in January 2020. Lux’s portfolio now has defense startups in “systems—space, air, sea, and land—that are involved in everything hardware and software,” Wolfe says. Although he makes clear that Thomas is honoring the ethics requirements that prevent a recently retired official from actively approaching anyone at the DOD, “Tony has been good in saying this is the person that you’re going to want to talk to.” This kind of effort, Wolfe notes, pales against the “ungodly amounts” that prime military contractors spend on lobbying.
“The Navy hosts product demonstrations every two years. That’s not fast enough.”
Rich Jenkins, CEO of Saildrone, knows a lot about evaluating water drones, and the Navy’s biennial, four-day events cannot compare to the field testing he’s put his products through. In 2019, one of his Saildrones completed the first autonomous circumnavigation of Antarctica, traveling more than 13,000 miles in 196 days, enduring 50-foot waves, icebergs, and 80 mph winds. “They need to put systems out in the field for four to six months,” says the sandy-haired Brit.
Steering through the DOD’s bureaucracy makes the high seas seem tame, and it may well be the single-largest impediment to the tech industry succeeding in transforming the defense industry. The barriers that tech startups face are structural, cultural, and all seemingly immutable. Looking around Saildrone’s headquarters—a huge former Naval Air station hangar in Alameda, California, fronted by a sprawling dock that leads directly to the waters of San Francisco Bay—it looks like a hardware startup. But each of these Saildrones of various sizes piled high, almost all of them the same bright safety orange, are a container for high-resolution optical cameras with machine-learning target detection as its devices patrol waterways to watch and listen in the right places.
“I look at major weapons systems and platforms as hardware enabled, yet software defined,” says Ellen Lord, who served as undersecretary of defense for acquisition and sustainment—the job DIU’s Brown had been nominated for—during the Trump administration. Lord, who had previously been CEO of the systems division of Textron, the $16 billion defense conglomerate, brought more of a tech-world perspective to the problems that have bedeviled Silicon Valley—namely, that for decades the DOD has bought software using the same two-years-in-advance schedule and funding mechanisms designed for hardware.
During her tenure, Lord enlisted the likes of Founders Fund partner Trae Stephens to study current methods of funding and acquiring different types of software, then offer a set of recommendations for improving those regulations. Congress has now given the DOD the ability through the National Defense Authorization Act to implement some of them. But that’s just the first step. “It’s an enormous amount of work to translate between statute and business practice,” says Lord, who is now a senior adviser at the Chertoff Group, a security risk management firm, “and then another enormous step to train the contracting and program management workforce to understand what they are allowed to do.”
Even with these incremental improvements, the system remains tilted by some of capitalism’s worst habits. In the past two decades, according to a February report by Open Secrets, defense companies have directed $285 million in campaign contributions and $2.5 billion in lobbying spending to influence defense policy. They’ve hired more than 200 lobbyists who have worked in the bureaucracy that regulates the industry and decides its funding. “How do you wrestle [programs] away from the large defense contractors who just spend billions on lobbyists in D.C.?” Jenkins says.
Wolfe, the venture capitalist, has a possible answer, one befitting his “expect the worst” outlook. He says that his firm, Lux, “and a handful of other venture firms and tech companies are all pushing. It’s going to take time, and, ultimately, probably a shock from a competitor [China or Russia] that really breaks down those walls.”
“[We’re] using China as the bogeyman, replacing the Soviet Union. Everything revolves around the defense of Taiwan.”
Chris Miller, who served as acting secretary of defense for President Trump from November 2020 to January 2021, is not exactly a dove when it comes to China. But he believes that post-Afghanistan, the establishment now needs a “great power competition” with China to justify the $700 billion-plus annual defense budget. “All they’re doing, again, is creating this narrative that somehow we’re at existential risk. No matter what happens, we’re going to be set with this huge, bloated military that’s focused on really exquisite, high-cost, very difficult to maintain, very technologically advanced weapons systems that are tied to a flawed war-fighting concept.”
Silicon Valley views China as its chief rival as well. One-fifth of the world’s trade—which is increasingly being ordered online and procured via tech companies digitizing supply-chain logistics—flows through the patch of Pacific Ocean between China’s southern coast, Vietnam, Malaysia, and the Philippines known as the South China Sea. Most of the world’s computer chips are produced in the region. One Taiwanese company in particular, TSMC, currently manufactures the vast majority of the advanced processors that control many aspects of Western life.
Although a pivot to China emerged as the focal point of America’s defense and economic priorities in the Obama era, concerns in the tech world ratcheted up in 2017, when the State Council of China’s New Generation Artificial Intelligence Development Plan (AIDP) laid out a strategy for achieving worldwide dominance in AI by 2030. “The world’s major developed countries are taking the development of AI as a major strategy to enhance national competitiveness and protect national security,” the report stated. AIDP calls on private companies such as Alibaba and Baidu to develop and apply AI to various social challenges—defense among them. At the time, China’s reported annual defense budget was $151.4 billion, up 7% from 2016, though it represented just a quarter of U.S. military spending that same year. (It grew to $209 billion in 2021.)
Silicon Valley is a little envious of China’s military-civil fusion (as the U.S. State Department officially characterizes it) and frustrated anew with the American status quo. “I don’t want their totalitarian system,” Brown says, “but no question they could move a lot faster with their chain of command than we can in our democratic one.”
Brown believes that China’s tech and economic ambitions are fueling this competition and inspiring more tech people to work with the DOD. “What [China] wants to do is supplant all of the technology leaders in every category—AI, cyber, autonomous systems,” he says. “They’ve already done that on the small side with DJI [in drones]. They want Huawei, Baidu to be the technology names of the future—setting the standards, leading global market share.” That kind of shift could have deleterious consequences for tech companies long used to attracting the world’s best talent and commanding the highest valuations, and it could impact America’s ability to use its tech sector to project its power and values.
But it is not a fait accompli. Defense tech’s amplification of tensions could have negative consequences for U.S.-China relations. “The threat here is not of China ‘taking over the world,’ ” says Jessica J. Lee, a senior research fellow in the East Asia program at the Quincy Institute for Responsible Statecraft, a nonpartisan foreign-policy think tank backed in part by both Charles Koch and George Soros. “[It’s] more of a worsening security dilemma leading to increasingly zero-sum, worst-case forms of competition that harm all countries and increase the chance of crises and conflict.”
“During the Cold War, the Secretary of Defense office would call Kelly Johnson at Lockheed Skunk Works and say, ‘We need to build stealth fighters.’ Johnson would say, ‘We know how. Give me $30 million. We’ll build one that will fly in 18 months.’ “
Trae Stephens, the Founders Fund partner who also cofounded Anduril, wistfully recalls the moment in 1955 when the nation’s preeminent aircraft maker and the DOD could collaborate at the speed of sound. For Stephens, it’s that spirit that needs to be rekindled. “That involved a level of trust and belief,” he says. “It would not happen today.”
If “the rules were gone and I could do anything I wanted,” Stephens says, he’d stop flinging R&D pennies at thousands of startups. Instead, he’d “focus on funding a smaller amount of companies with significant financial commitments that are truly going to advance the mission of the war fighter.”
He’s not alone in thinking that the DOD’s broad efforts are largely performative. “All the services are building out their little innovation labs, which is cute, and it works for a while, and it’s necessary,” says Miller, who was acting secretary of defense less than a year ago. “When they go up on the Hill, the acquisitions chiefs, the service secretaries, and service chiefs can talk about how innovative they’re being. But when you follow the money, it’s just not there.” The current DOD effectively agrees. “You’ll be delighted to know that one of the things we’re looking at is all those quote-unquote innovation organizations that have sprung up everywhere,” says Heidi Shyu, undersecretary of defense for research and engineering, who wants to understand precisely how much tech has transitioned into production.
Whether the DOD continues to pursue a broad investment strategy in startups or places a few big bets, a Silicon Valley-centric defense industry may well move faster than the handful of legacy contractors who dominate it today. The tech players backing this future know it may be a slow build, but just as COVID-19 accelerated telehealth and remote learning, boosting fortunes, so too could a shock like a military conflagration.
But that speed comes with costs. Like jobs. Software programs require fewer workers than building ships, for example. The engineers writing those algorithms are also likely to be concentrated in a few urban hubs and share a social class. “There’s a handful of jobs being created, but they’re going to upper-class, well-educated elites who have master’s degrees or PhDs” in fields such as machine learning, says Yale historian Michael Brenes, whose 2020 book, For Might and Right, chronicles the way in which Cold War defense spending reshaped America.
If tech’s urgency to scale its businesses at venture speed translates into rushing AI systems into the battle space, that could be disastrous. “We don’t understand yet exactly when and why they fail,” says Chris Meserole, director of research of the Artificial Intelligence and Emerging Technology Initiative at the Brookings Institution, the center-left think tank. “Until we do, it’s going to be very hard to understand when and why we should use them.”
Even if AI succeeds, there will be implications. Today’s standard for the armed services is that in order for an autonomous weapon to be used predictably and ethically, a human decision-maker must always approve its actions. But as these systems become more advanced, the human factor could become the thing that impedes its performance. The temptation to remove this weakest link might be too great. If one side does it, the other side may be forced to follow suit. That same quandary might spread from one kind of weapons system to the next. What happens when no human operators are left?
Some in the tech community believe that using AI in warfare will produce a cleaner, more precise form of aggression that surgically strikes at enemies while limiting “collateral” damage to others. But a September report from Airwars, a U.K.-based nonprofit that monitors civil harm in conflict, presented a range of at least 22,679 and possibly as many as 48,308 civilian casualties from U.S. targeted air or drone strikes during the war on terror in Afghanistan, Iraq, Libya, Pakistan, Somalia, Syria, and Yemen. (This would not include the event in August, which killed 10 civilians, including 7 children, in Afghanistan.) Airwars admits that these numbers are conservative. “We are therefore looking at a fraction of the overall civilian harm in these countries,” the organization states.
As a result, there’s a growing movement to restrict the use of AI-powered weapons. The nonprofit Human Rights Watch cofounded the Campaign to Stop Killer Robots, a coalition of 180 like-minded groups in 65 countries calling for an international treaty to ban fully autonomous weapons. “The techno optimism that infuses Silicon Valley and of course the Department of Defense is just the way it’s always been, unfortunately,” says Mary Wareham, advocacy director of HRW’s arms division. “We’re sleepwalking into a really grim future if all we’re going to talk about are the benefits of technology, the promises that it will bring, and not deal with the downsides, the dangers, and the people who are going to be threatened by such applications of tech.”
Thirty-six countries have endorsed the goal of a treaty banning autonomous weapons, which has the support of UN Secretary-General António Guterres, who has called autonomous warfare “morally repugnant and politically unacceptable.” China supports an international treaty ban, though it wants only to limit their use, not their creation in the first place. The United States has rejected any agreement on lethal autonomous weapons, calling it “premature.”