America won the Second World War with mass production, logistics, and technology. The much-vaunted German Army ran on horses and hay; the U.S. Army—not to mention the Red Army we supplied—ran on the deuce and a half and the Jeep.Two of the three major Axis powers never built an aircraft carrier. The U.S. Navy built 151 andhad enough left over for a fleet of refrigerated ice cream barges.
The idea of America as the Arsenal of Democracy in World War II—innovative, productive, and hard-working—is now firmly a part of the story we tell about ourselves. It is a source of pride, patriotism, and inspiration. And it’s a true story.
But that was then, and this is now.
Today, Americans are waking up to the reality that we can’t make things in sufficient quantities to keep us safe, while our principal adversary is flooding the world with its manufactures. Many of our allies are in even worse shape.
Rebuilding and rearming before it’s too late are the urgent tasks before us. Thankfully, industrial decline is a solvable problem—and a choice. We can make better choices, create better incentives, and use the strengths we still have to our advantage.
Not only is mass production a solvable problem. It’s already being solved. Half of our business at Palantir Technologies, where I work as chief technology officer, is with commercial clients. We work every day to create software that helps the free world’s most important industrial concerns supercharge production, see into their supply chains, and empower their workers, from the C-suite to the assembly line.
I’ve seen the results, and I believe this model—what’s sometimes called software-defined production—is key to rebuilding the American industrial base.
No More Ice Cream Barges
By now, readers will be familiar with the basic contours of America’s production problem and the world’s reliance on Chinese manufacturing. But it’s worth recapping, to remind ourselves of how serious and severe the challenge really is.
In the 1990s, a Western executive said that thinking about the size of the Chinese consumer market was “like trying to think about the limits of space.” Something similar could be said about Chinese manufacturing, after three decades of industrial policy, investment, and learning by doing.
China today has a nearly one-third share of global manufacturing output—not as much as the United States at the height of its postwar power, but a massive amount nonetheless. It was America’s manufacturing dominance that enabled the spread of our commerce and power across the globe in the nineteenth and twentieth centuries. China’s vast industrial base gives it the same opportunity.
China has nearly half of global shipbuilding capacity. It is using this capacity to build drone aircraft carriers, large LNG tankers, and roll-on/roll-off ships for auto exports. China has 232 times more shipbuilding capacity than the United States, whose industry has consolidated to the point where we have to choose between building submarines for our allies or for ourselves. We don’t have the luxury of ice cream ships anymore.
It’s not just ships. Two companies in Shenzhen make virtually all of the world’s commercial drones, while the United States barely has a commercial drone industry. China surpassed the United States in share of global semiconductor production within the last few years, and appears poised to clean up in so-called legacy chips that power commercial electronics, weapons, and much more. China is also the world’s largest producer and exporter of automobiles, gas-powered and electric. The United States still has an impressive auto industry, but the number of vehicles assembled here hasn’t changed much since the turn of the century. U.S. automakers now face an existential threat as Chinese competitors like BYD build transplant factories in Latin America, Southeast Asia, and Europe.
These aren’t random examples. All three technologies—drones, chips, and cars—were invented in the United States (in the case of cars, their mass production was invented here). General Atomics, Intel, and Ford Motor Company were pioneers. In the span of a lifetime, the United States went from dominating production of all three to facing the fight of our life in these industries. Why?
America’s crisis of production is ultimately a crisis of productivity. Total factor productivity has stagnated during the past half century, diverging from trend in the 1970s. Manufacturing productivity did better for longer thanks to a booming electronics sector, but since the Great Recession, it too has stagnated. This stagnation means American factories are missing out on an era of tremendous automation and growth. The “alien dreadnoughts” are being built. But they are being built on the other side of the world.
This “Great Stagnation” has many causes. Until recently, manufacturing hasn’t received much love from investors, who preferred capital-light businesses that promised easier returns on their money. The talent has followed the money. We’ve seen the best minds of our generation sucked into a SaaS hole of darkness, designing targeted ads for banal products. Meanwhile, the manufacturing workforce, white- and blue-collar, has aged and shrunk. Fewer factories mean fewer line workers and managers with the knowledge to build more factories.
Red tape, bad incentives, and apathy have made matters worse. This is apparent in the market for defense technology, which is not a functioning market. The Pentagon is the sole buyer for defense products, and has leaned into its monopsony instead of encouraging market forces. This has resulted in consolidation and conformity from industry. Only recently has America woken up to the reality that this consolidation has hurt competition, innovation, and industrial capacity. A lot of great American companies were folded into their competitors. But much more importantly, many more companies have exited defense altogether, a direct result of monopsony with onerous rules, emphasis on costs over value, and an inability to reward speed. Contractors that survived the winnowing have become extensions and wards of the state, content to earn small profits on contracts while costs balloon and the commercial world passes them by. Companies that don’t fit the mold, like the new production champions we work with, are frustrated by byzantine rules and contracts that don’t reflect the way America does business.
A while back, Palantir partnered with a major American defense contractor to improve its performance on a standard cost-plus shipbuilding contract. The pilot program was a success. The contractor and its employees loved our system. But they didn’t buy it. Why? Because of the incentives baked into their contract. They’d have to give back any money they saved using our software on recompete.
This was a hard lesson in how broken DoD contracting can be. Instead of rewarding companies for being lean, mean, fighting machines, the DoD penalizes them. Should we be surprised that the United States is an also-ran in shipbuilding, when its most important customer encourages bloat and complacency?
Above all, America suffers from a lack of leadership and vision. Our prosperity, paired with the cinematic ending to the Cold War, bred overconfidence and passivity. Too many leaders in business and government fooled themselves into believing that our system had the Mandate of Heaven and that our economy was a money-minting machine that effectively ran on autopilot. “Industrial policy” became a four-letter word, treated as a relic of an economic model that had fallen alongside the Berlin Wall. In Soviet fashion, our own history was revised to downplay instances where state investment and direction had incubated new technologies. Counterproductive DoD contracting models may have survived, but many more effective programs faded away.
Even as supply chains internationalized and manufacturing shifted offshore, our leaders consoled themselves that America would still make the “important” things. And they believed that factories that were quick to go could come back just as quickly. If we ever needed manufacturing, we could flip a switch and turn the arsenal back on. Ice cream ships for everyone. Our struggle to supply Ukraine in its fight for survival has shattered that delusion.
Contrast this passivity with the incredible activity of our enemies. Xi Jinping and Vladimir Putin are ruthless dictators, but whatever else can be said about them, they aren’t cheap-seat observers of history. They’re empire builders with grand visions of what they hope to achieve, and they are mobilizing their civilizations to make those visions reality.
In Silicon Valley, we have a word for this type of leader: “founder.” We can deplore their methods and their vision, but we underestimate them at our peril. Plenty of founders fail, but not all. Some of them change the world. The ones that do usually aren’t known for the friends they made along the way.
Give America her due. We are a free country blessed with an extraordinary inheritance of ideas, riches, and resources. We have strong leads in many fields, including software engineering, computer science, and product design. But we have lost mass production—and we’ve given it to a mortal foe.
This is a grave problem in a dangerous world. After all, we can’t shoot software. Our brilliant designs won’t count for much if we can’t produce them in numbers sufficient to make a difference.
The defining story of this decade has been the realization by Western elites that hardware matters. But it would be a mistake to conclude from this story that software doesn’t matter, or that the strengths we have are irrelevant to the problem at hand. The opposite is true. Modern industry is all about building production around software so that human actors can manipulate machines at speed and scale.
If we’re going to make things in America again, we’ll have to play to our strengths. That means understanding how the software we’ve pioneered can be used in industrial processes—and using it to build, before our adversaries steal even this advantage from under us.
Software-Defined Production
Palantir won its spurs providing software to the U.S. military and intelligence community after 9/11. The challenges we faced then were immense. American troops were operating in some of the most forlorn parts of the world. The military’s IT infrastructure was fragmented by distance, combatant commands, services, air-gapped environments, and countless legacy systems that didn’t talk to each other. We built software on top of that infrastructure, fusing disparate systems so that analysts in the Pentagon and troops in Parwan Province could pierce the fog of war and close kill chains.
Modern industrial systems have their own fog of war. They are staggering in their complexity. The average commercial passenger jet is composed of millions of parts, spread across factories, companies, and countries. Inevitably, some of the parts are defective. Others don’t arrive. Mistakes are made on the assembly line. Some are logged and addressed. Others aren’t. The assembly line moves anyway and problems pile up. As this frenzy occurs, board members and activist shareholders are breathing down the necks of corporate executives, asking why production targets aren’t being met. Putting together planes under these conditions is a feat of intense coordination, logistics, and willpower, the industrial equivalent of D-Day.
Software has the power to make this chaos worse, if what it shows on the screen bears little resemblance to what’s happening on the ground. Corporate leaders may think they’re in control of the plane—they can see the charts with their own eyes, and the line is going up!—but in reality, they’re turning a toy wheel. Their network is fragmented across suppliers, factories, and machines, with little coordination. Inventory that managers think is in the warehouse isn’t there. Quality checks they think occurred never did. And all along, the glowing screen seduces them into thinking they’re really in control—until the door plug blows out on the plane.
There is a better way. What if companies could create a superstructure of software over the entire, fragmented production process, just like we did for the government? This software would absorb and analyze data from the company’s countless suppliers, components, machines, and workers to create a complete model of the production process. And it would control physical machines, like industrial robots and machine tools, allowing refinements to the production process on the fly.
This is the promise of software-defined production. It allows managers to regain control of sprawling bureaucracies. It helps workers understand how to act amid complexity. It connects strategy with operations. It acts as a digital aid to human agency, allowing them to execute and win.
Just a few decades ago, modeling, analyzing, and controlling an entire production process digitally was the stuff of science fiction. Now, thanks to advances in computing power and sensors, it is reality.
We put it into practice in 2015 when Airbus scaled production of the A350, a wide-body, dual-aisle passenger jet. Airbus’s CEO had a problem. He had pledged to investors that Airbus would make fifty A350s in the first year of production. Halfway through the year, they were on track to make sixteen.
Palantir’s Foundry platform helped Airbus understand and reengineer its production process. It fused data on schedules, shifts, parts, deliveries, defects, and much more into a single platform. Insights from this system helped Airbus reduce defects, prevent accidents, and respond flexibly to supplier delays. When the deadline arrived, Airbus made forty-nine planes, just one shy of the target. A fundamental tenet of military strategy is that no plan survives first contact with the enemy. This one came close.
The success of software-defined production depends on the ability of the software to reflect and respond to reality. That requires physical proximity to production and boots on the ground.
When U.S. troops deployed to the Middle East, Palantir engineers deployed with them to understand the challenges they faced and ship code that reflected conditions at the front. We call this “forward-deployed engineering,” and when we first proposed it, investors balked. They said we were wasting money adding an expensive client-services segment to our business. They were wrong. Because we showed up, we could get to the root cause of the problem and tailor systems to the needs of troops at the tip of the spear. The best testimonials to our success come from the troops who insisted that Palantir accompany them into battle. I have many such stories.
We operate the same way for our industrial customers. Some of the most impressive results of software-defined production occur through the collaboration of engineers and assembly workers on the factory floor.
We saw this in Airbus’s Hamburg factory, when an assembly worker was recovering from surgery. She couldn’t operate heavy machinery, but she could operate a computer. She used her knowledge of Airbus production to analyze common defects, where they occurred, and the delays they caused. This was forward-deployed engineering at its best. An ordinary worker used powerful data-analysis tools to validate what her gut was telling her about what was going wrong on the line. The whole company was more productive as a result.
Software platforms can also train workers who lack the skills for complex production. Skilled labor is a bottleneck to the expansion of American manufacturing. As TSMC is learning, there’s a finite domestic supply of construction workers, technicians, and managers who can build the world’s most advanced factories in the desert. This shortage is downstream of the decades-long decline of our industrial ecosystem. Tim Cook put it starkly in 2017: “In the U.S., you could have a meeting of tooling engineers and I’m not sure we could fill the room. In China, you could fill multiple football fields.”
Here, too, software can hack the process and train workers faster. Panasonic Energy makes EV batteries at several factories in the United States. The company needed to scale its workforce of maintenance technicians, so they turned to us for help. We created an AI copilot for every technician that enables a high school graduate to operate with the knowledge and experience of Panasonic’s top Japanese experts.
This tool has shortened the runway to train Panasonic’s maintenance techs from as long as six months to a few weeks. It allows Panasonic to take workers with zero exposure to manufacturing and re-skill them for rewarding work in a strategically important sector.
Technology is no longer the province of West Coast coders and cyberspace fantasies. It is fundamental to “blue-collar” work, learning by doing, and a mass manufacturing revival.
Ultimately, the best evidence in favor of software-defined production is that the very countries that dominate manufacturing are trying hard to incorporate it. One of the key objectives of Beijing’s Made in China 2025 strategy is the digital upgrading of manufacturing. Huawei’s gargantuan 5G deployment is connecting heavily automated factories with heavily automated container terminals like the one at Shanghai to boost exports. Chinese Communist Party officials frequently jawbone traditional industries for being slow to upgrade. The CCP knows that digitization and automation are the only ways to sustain China’s breakneck pace of production as its working-age population craters in the decades ahead.
China knows that software-defined production is the future. They want to get there first and do it better.
We Have the Technology
Fixing our production problem will require urgency and vision to match our adversaries. It will require founders. Perverse incentives in defense procurement have to be eliminated—or at least circumvented through innovative programs like Replicator. Soul-sucking, project-delaying regulations will have to be slashed. Industrial plants will have to be expanded and recapitalized, by private and public sources. And then the hard work of production has to happen.
This is a daunting challenge, but to paraphrase the Six Million Dollar Man, we can rebuild the American industrial base. We have the technology.
Earlier this year, Secretary of the Navy Carlos Del Toro traveled to Korea and came face-to-face with the East Asian development model. He was there to drum up interest for investments in America’s shipyards. Among his stops was the nearly two-thousand-acre Ulsan shipyard of HD Hyundai Heavy Industries. Ulsan is the world’s largest shipyard, operated by the world’s most prolific shipbuilder, in a country with a famously aggressive industrial policy. Del Toro was impressed by what he saw. “I could not be more excited at the prospect of these companies bringing their expertise, their technology, and their cutting-edge best practices to American shores,” he said.
The United States has a lot to learn from HD Hyundai’s expertise and best practices, to be sure. But on technology, Del Toro was only partly correct, because America is already providing some of the technology that dazzled him at Ulsan. HD Hyundai uses Palantir Foundry to design ships, improve quality control, and reduce on-site accidents. The partnership is already showing teeth. We’re codeveloping an unmanned surface vessel powered by AI that may one day help the navies of the Free World reconnoiter and defend the Pacific.
American technology is already the backbone of some of the world’s most impressive industrial giants. If we’re going to revitalize the Arsenal of Democracy, we need to unleash the technology at our fingertips.
