Rebuild
the arsenal of democracy
America’s defense industrial base, which has helped to deter war and maintain peace for decades, has failed to keep pace and adapt with the times. It is being left behind by evolving great power threats, technological innovations in areas such as low-cost robotics and artificial intelligence, and a commercial manufacturing revolution that is enabling the hyper-scale production of everything from self-driving cars to reusable rockets to intelligent devices of all kinds. America and our allies increasingly lack the industrial capacity to deter — and, if necessary, fight and win — a great power conflict.
This problem has been long in the making. For decades, U.S. and allied military power has been defined by ever-smaller numbers of exquisite, expensive platforms and weapons that are slow, difficult, and costly to manufacture. Our stockpile of critical munitions has taken years to produce and would take just as long to replace. And yet, years of war games suggest the U.S. military would run out of these weapons in less than one week of a war with China. This is not hard to believe after Ukraine expended a decade’s worth of U.S.-produced tactical weapons in just the initial months of combat with Russia.
At the same time, the survivability and enduring relevance of traditional military systems are being called into question as never before by the proliferation of AI-enabled weapons, autonomous vehicles, ubiquitous sensors, modern software, and other advanced capabilities that are lower cost and and mass producible. This is not a distant concern. It is visible on recent battlefields in Nagorno-Karabakh, Iraq and Syria, Israel and Gaza, and Ukraine. The future of warfare is already here.
A great power conflict would be a war of production, attrition, and regeneration at scales that are nearly unfathomable to our defense industrial base. If such a conflict were to occur, we may be ready for day one, but we are utterly unprepared for day 30, let alone day 300. America and our allies need to rebuild the arsenal of democracy, and we believe that is possible.
We believe the demand for a new approach to defense production has never been higher among U.S. and allied leaders.
We believe our defense industrial base must be capable of producing orders of magnitude more weapons and military platforms than is now achievable with our traditional defense systems and their equally traditional means of production.
We believe that future deterrence depends on the mass production of new classes of autonomous vehicles and weapons — not just small tactical systems, but larger, long-range capabilities that create real advantages in high-end warfare.
We believe that what passes for large-scale production in national defense today is actually unrecognizably small by the standard of modern commercial manufacturing, which is the standard we must achieve to prevail in the future of warfare.
We believe the technology, resources, and human capital to forge a different kind of defense industrial base already exists in the United States and allied countries, and they can be mobilized to launch a new golden age of defense manufacturing.
We believe Anduril has a unique role to play, so we are doing what we have consistently done before: We are moving out rapidly and spending our own money based on our own convictions and the clear intent of our government partners. We have recently made major investments to scale our production capacity for loitering munitions in Georgia, solid rocket motors in Mississippi, large autonomous undersea vehicles in Rhode Island, and even larger undersea vehicles in Sydney, Australia. These investments are providing Anduril with significant manufacturing capability across a range of products. But we see the need to go even further. We need to move from large-scale production to hyper-scale production.
This requires a new product, and we are making an even larger investment of our own money to deliver it. This product is not a weapon itself, but a maker of weapons. It is a software-defined manufacturing platform to produce advanced defense capabilities at hyper-scale. It has been years in the making, and it will make possible everything we will build in the future.
We call it Arsenal.
the problem of defense production
The challenges plaguing our defense industrial base have taken decades to materialize, and they stem from how U.S. and allied governments have defined their requirements for military power. With the end of the Cold War, it appeared that the era of great power competition was over, and that no global threat existed to our collective military primacy. We could not imagine losing large numbers of weapons and military platforms in combat, so we believed we could afford to build militaries that were effectively irreplaceable — forces composed of smaller and smaller numbers of expensive, exquisite systems that would be so stealthy, so survivable, so technologically superior that they would never need to be mass manufactured.
The result is that our current weapons and military platforms are defined by exquisiteness. They are exquisite by design, produced in such small numbers and modernized so rarely that government feels compelled to add every conceivable requirement, lest they miss a generational chance to do so. They are time-consuming to develop, often taking years of painstaking effort and small fortunes of R&D funding that are borne nearly exclusively by taxpayers not corporations. They are difficult to build and nearly impossible to build quickly or at large scale, requiring highly specialized labor, rare materials, secure facilities, manual production processes, customized supply chains, and other unique elements. They are also difficult to the point of impossible to change or update quickly with new technologies, which only makes manufacturing harder, because weapons are filled with obsolete components that few companies are able, or even willing, to build anymore.
This does not happen by accident. Though rarely stated, exquisiteness in traditional military systems is actually encouraged and rewarded. Government has long assumed that more exquisite materials, components, and manufacturing processes are the only way to meet the military’s unique requirements for durability, security, and effectiveness — even though it makes those products significantly more expensive and less producible. Traditional defense industry also prefers complex, hard-to-build products for different reasons: They reduce competition by raising the barrier to entry, and the increased cost, time, and labor required to manufacture more exquisite products are simply charged to the government on cost-plus contracts. Government often makes all of these problems worse through slow, onerous verification, certification, and testing burdens.
Such exquisite, costly, and slow manufacturing processes do not lend themselves to large volumes of output. In Fiscal Year 2023, for example, the United States only planned to produce one or two submarines, several warships, 22 tanks, and a few dozen advanced fighter jets. Annual full-rate production of advanced systems such as long-range bombers or intelligence satellites can each be counted on one hand. The result is that the U.S. military is literally shrinking. Each year we are retiring ships, combat aircraft, and other major platforms at a faster rate than the industrial base is capable of replacing them.
The story is no better when it comes to weapons, especially the critical munitions that would be required to deter or prevail in a great power war. In Fiscal Year 2024, for example, the United States only plans to produce [27] Long-Range Anti-Ship Missiles (LRASM), [XX] Standard Missile 3s, [XX] Standard Missile 6s, [55] Tomahawk cruise missiles, and [535] Joint Air-to-Surface Standoff Missiles - Extended Range (JASSM-ER). Most of these weapons cost several million dollars apiece, and the capacity to scale production does not exist. It is not hard to imagine how the United States would run out of these and other critical munitions in a matter of days in a war against China and then struggle to replace them on a relevant timeline. Indeed, even the less taxing burden of rearming Ukraine is proving unbearable for our defense industrial base.
The deeper problem is that this legacy approach to defense production does not scale. Even extraordinary commitments of national will and resources, similar to World War II or the late Cold War, would not result in rapid and significant increases in defenses production. In their designs, materials, components, labor forces, supply chains, manufacturing processes, and more, current military platforms and weapons are more exquisite than earlier generations of the same systems in almost every way. They are small-batch, hand-crafted, artisanal products that are effectively impossible to produce and replace at high volumes, even with increased spending. Indeed, when Ukrainian soldiers depleted the U.S. inventory of Stinger missiles in just several months of fighting with Russia, it was estimated that new systems would not reach the battlefield for two to three years, and even that required bringing workers out of retirement. No one else knew how to produce them.
Our reliance upon the military equivalent of luxury goods was more defensible in a world of U.S. primacy and no great power threats to it. But that world is gone. The Chinese Communist Party has emerged as a strategic rival, and it is embarked upon an unprecedented military build up that is specifically designed to achieve what we have long believed to be impossible: to attrit the centerpieces of U.S. military power and force us to reconstitute our effectively irreplaceable weapon systems.
China does not really have that problem. Its military expansion is being powered by a domestic manufacturing engine that is outproducing U.S. industry in everything from warships and combat aircraft to missiles and drones. This did not just happen overnight either. It was the result of long-standing U.S. policy toward Beijing that aspired to forge a symbiotic bilateral relationship with America as buyer and China as builder. As a result, for four decades, the United States has systematically de-industrialized while China has hyper-industrialized, and the resulting disparity in manufacturing capacity is colossal.
Take just one example: It is estimated that China possesses more than two hundred times the shipbuilding capacity of the United States. It accounts for close to half of the entire world’s ship production, while the United States makes up less than one tenth of one percent. A similar story could be told across every other segment of military production.
The United States and our allies still need the exquisite military systems that we have relied upon for so long, as well as the industrial base that produces them. But this alone is insufficient. Our current defense industrial base cannot outproduce China. That would require massive amounts of time, money, and specialized labor — none of which we have.
The number of weapons we actually need to deter great power war is an order of magnitude more than we can build at present. We need a different approach to generating military power, and this is now possible.
a manufacturing renaissance
In 2013, Tesla’s production goals had begun to amuse Wall Street. To justify the company’s $120 share price, equity analysts at Bank of America and Merrill Lynch believed that Tesla would need to sell 321,000 cars in 2020 — more than fifteen times as many as the company projected for 2013. This meteoric increase, the analysts wrote, represented “300,000 reasons” to short Tesla stock. They noted that Tesla’s planned rate of production growth was historically unprecedented in the automotive industry. Thus it was believed to be impossible. Even as the company scaled, the former Vice Chairman of General Motors, Bob Lutz, predicted in 2017 that “Tesla is going out of business.” He gave the company two years to live.
By 2020, not only was Tesla still alive; it had given Wall Street analysts 500,000 reasons to eat their words. Despite a global pandemic that shut down large parts of the world economy, Tesla produced 50 percent more cars than the original target that Bank of America had considered so preposterous. What the critics failed to understand was that the traditional metrics of the automotive industry did not apply to Tesla, because it was not designing and building cars in the same way that its competitors did. Tesla was pioneering, in the words of one Toyota executive, “a whole different manufacturing philosophy.”
What were the hallmarks of this new approach? In short, Tesla used software to revolutionize hardware. It broke with the traditional automotive industry practice of treating the vehicle’s operating software as a secondary concern — a practice that Ford CEO Jim Farley described in 2023. ‘[I]t’s so difficult for legacy car companies to get software right,“ he explained:
“We farmed out all the modules that control the vehicles to our suppliers because we could bid them against each other. So Bosch would do the body control module, someone else would do the seat control module, someone else could do the engine control module.… The problem is, the software is all written by 150 different companies, and they don’t talk to each other.… And we can’t even understand it all. So that’s why at Ford we’ve decided in the second-generation product to completely insource the electric architecture. And to do that, you need to write all the software yourself. But just remember, car companies haven’t written software like this. Ever.”
Well, not all car companies. This is exactly what Tesla did more than a decade earlier: It treated the vehicle’s software, not the vehicle itself, as the actual platform and built that software on its own as one integrated piece of technology.
Tesla is not unique in this way. This is how Apple revolutionized mobile devices and other consumer electronics. It is how Nvidia built a new generation of supercomputers. It is how SpaceX pioneered reusable space launch vehicles and built and launched more satellites in several years than all of humanity had prior to that point. These and other leading commercial companies are achieving what many thought impossible because they are, first and foremost, software companies, and it is software that enables them to design, develop, and manufacture their hardware products in entirely new and different ways.
With Tesla, its software-first approach gives the company greater control over its hardware architecture, which it uses not only to build a fully electric vehicle, but also to do more mundane but equally important things. Tesla is able to simplify its design by consistently reducing the vehicle’s overall number of parts, moving to fewer software-controlled components and cutting other hardware modules altogether. Its open software architecture enables Tesla to replace vehicle subsystems faster and more easily with cheaper, more commercially available alternatives with more diverse supply chains. By prioritizing software, the company is able to iteratively redesign its hardware at every level at speeds that were unheard of in the traditional automotive industry, and all of this radically decreases costs and increases producibility.
Tesla’s software-first approach also extends into the manufacturing process itself and enables the company to break many of the old rules of production. Traditional automotive manufacturing was linear. One team’s ability to make progress was gated by every preceding team in the production process finishing its work. Tesla uses a common suite of software tools to build a digital model of its vehicles. This single source of truth liberates each engineering team to change its respective part of the vehicle as much or as often as necessary without fear that it would disrupt another team’s work.
As a result, Tesla has been able to move rapidly into large-scale production and then iterate nonstop on every aspect of the vehicle as if it were still in development. Indeed, at its current peaks of production, Tesla is introducing hundreds of changes and new features into live manufacturing lines [each week] in order to make its vehicles easier, faster, and cheaper to build. This is how Tesla achieved hyper-scale production — a 23-fold increase in its output of vehicles — in only seven years.
In an attempt to learn new lessons from commercial production, the defense enterprise has recently become enamored of concepts such as digital engineering, advanced manufacturing, and “Industry 4.0” — so much so that these ideas are often reduced to buzzwords or corporate initiatives that have become, in the words of Secretary of the Air Force Frank Kendall, “overhyped.” There are certainly places in government and traditional industry where new approaches to manufacturing are being used to good effect, but if they are only used to repackage old ways of working — if they are not used, as they should be, to enable people to work differently, transform organizational culture, and fundamentally alter how products are designed and built at scale — then the while thing misses the point. It becomes akin to electrifying the horse cavalry.
designing
for simplicity
& scale
Solving the real problem of defense production requires a more radical approach. To be able to win, and thus to deter, a great power conflict, the United States and our allies require an order of magnitude more weapons and military platforms. Our current inventory is literally missing a zero. Producing at this scale is impractical with traditional military systems and their equally traditional means of production. We simply do not have the money, time, and specialized labor it would take.
What is possible, however, is supplementing our current force with alternative military capabilities — the kinds of lower-cost, mass producible, more autonomous vehicles and weapons that are changing the character of warfare on contemporary battlefields in Ukraine and beyond. Indeed, this is the only way the United Staters and our allies can generate the amount of military power we need in the limited time we have to build it. These nontraditional military systems must be produced in equally nontraditional ways, and that is exactly what the commercial manufacturing revolution now makes possible.
This requires new thinking as much as new technologies or new tools. Our major choices must be diametrically opposed to those that characterize traditional defense programs, and this must happen, first and foremost, at the level of product design. We must design products that are simple and producible, not exquisite and irreplaceable. We must design products that maximize the use of low cost and widely accessible components, not highly expensive and specialized ones. We must design products that intentionally take advantage of robust commercial supply chains and reduce dependencies on narrow, defense-specific supply chains. We must design products that require as little specialized labor as possible and instead are easy, fast, and cheap to manufacture by the largest possible workforce. If these and other design choices like them are not made deliberately upfront, it is too late. No amount of digital engineering or other novel tools applied later in the manufacturing process will overcome the inherent limitations to mass producibility that were designed in at the outset. This is why the commercial manufacturing revolution will simply never apply to F-35s, Destroyers, or Tomahawk missiles.
Designing for simplicity and scale has characterized Anduril’s approach to manufacturing since our founding, and this is only possible because we put software first.
Our Lattice software platform gives us total control over the architecture of every hardware product we build. It enables us to compose and recompose our products quickly and easily with different sensors, computers, radios, and other mission systems to adapt to new threats or integrate new technologies. Lattice also delivers autonomy across every aspect of the missions that our products perform — how they sense and make sense of the operational environment, how they process and share information, and how they maneuver and collaborate with one another to achieve their objectives. This software-first approach enables us to make completely different hardware choices that not only maximize operational effectiveness but also make each system faster, easier, and cheaper to manufacture.
For example, traditional air defense systems are architected around human limitations: Because U.S. and allied militaries only have small numbers of specially trained air defenders who operate radars and other sensors, those sensors must be exquisite and highly capable over long distances, which makes them expensive and difficult to produce in large numbers. Lattice enabled us to take the opposite approach, built around machine capabilities not human limitations. Because the software processes, fuses, and exploits all of the sensor data without need for humans in the loop, it is possible to replace more expensive, less producible sensors with lower cost, commercially available alternatives that, while less capable than their traditional counterparts, can be deployed in larger numbers to greater operational effect. We took a similar approach with our Pulsar electronic warfare systems: The hardware is composed of low-cost, commercial components, but Lattice enables the system to autonomously interrogate the electromagnetic spectrum and deploy exquisite countermeasures.
More recently, we are scaling this software-centric approach to design simpler, more producible autonomous vehicles and weapons. For example, with our large and extra-large autonomous undersea vehicles, the Dive LD and XL, we jettisoned the traditional practice of making the entire hull one large pressure vessel and instead adopted a free-flooded architecture with smaller pressure vessels just to protect the vehicle’s mission computer, navigation system, and other critical components. This enabled us to replace a more expensive steel hull with a cheaper composite hull that could be 3D-printed. The result is a low-cost autonomous system that can be manufactured at larger scale and deployed in large numbers to greater effect.
We are taking a similar approach on Fury, our autonomous fighter jet that informs our approach to collaborative combat aircraft programs. From the start, we designed Fury around an aluminum rather than a titanium airframe: The latter is stronger, but the former is lighter, cheaper, and more available. We chose simple mechanical fasteners rather than time-consuming bonding processes for the aircraft’s assembly. We determined that existing, forged landing gear from other aircraft, though tempting to reuse, were overly complex for an attritable aircraft, so we chose to machine our own, simpler design. The opposite logic applied on propulsion, so we chose to source a mature engine from a well-known, domestic, commercial manufacturer. The resulting aircraft is designed for mass production. Its core components are sourced from a reliable, broad base of vendors, and it can be assembled by manufacturing technicians with little specialized training.
With all of our products, we are making deliberate choices to design them around commercially available components and production processes, because it is simply not possible to achieve the scale of weapons production required any other way. For example, on the Experimental Test Vehicle, the low-cost cruise missile we are building for the U.S. Air Force, we are not using the standard, expensive cast metal fuel tank with its more limited defense-specific supply chain, but instead are utilizing a cheaper component produced through the same commercial casting process used to make large children’s toys — a process that many vendors can do at large scales. We are also forgoing the exquisite composite production processes that are used to produce many weapon airframes and instead are using a heated pressing process used to make fiberglass bathtubs. The result is durable enough to meet requirements, but less expensive, more mass producible, with a far broader and more resilient supply chain. Indeed, it means that the many companies in the United States or allied countries that are capable of casting bathtubs can now be part of our supply chain to build and scale weapons production.
software-enabled manufacturing
Generating exponentially more weapons and military vehicles on rapid timelines, as we must, requires fundamental changes not only to how those systems are designed but also to how they are manufactured. Over the past several years, drawing upon the experiences of Tesla, Apple, SpaceX, and other leaders of the commercial manufacturing renaissance, Anduril has been developing a different approach to defense production. We have been proving and maturing this approach across a range of products — from weapons and solid rocket motors to autonomous aircraft and submarines — and a growing footprint of manufacturing locations that now includes California, Mississippi, Rhode Island, Georgia, and Sydney, Australia. Indeed, this different approach to manufacturing is a major reason why Anduril, despite our lack of “traditional“ defense production experience, is being selected to perform on major defense programs, such as the U.S. Air Force’s Collaborative Combat Aircraft and the Australian Navy’s Ghost Shark Extra-Large Autonomous Undersea Vehicle. In short, we are breaking many of the established rules of defense production, and here too, a software-first approach makes this possible.
Traditional defense manufacturing is a linear process that starkly differentiates between design, development, and production. Much of this is made worse on cost-plus contracts by acquisition policies that proscribe exactly how and when a product should move from one phase to the next, while requiring senior government approval to do so. This makes a product’s path to production slower and more expensive. It also introduces risks. A product in its prototyping phase and in its production phase is very different. At times it is effectively a different product altogether. It is only by beginning high-rate production that it is possible to learn how a product must be redesigned and redeveloped to make it more producible. Delaying these learnings only makes it more difficult, costly, and time-consuming to manufacture a product at large scale.
It is possible to overcome these old obstacles with software and adopt a more modular, non-linear approach to defense production. Because Lattice provides us greater control over the architecture of our products, we can make changes to any aspect of the product — from software applications and algorithms, to hardware subsystems, to the overall product itself — with limited risk or regret, regardless of whether that product is in prototyping or production. This enables us to accelerate minimum viable products into production and then continue to modify, fix, and improve them as if we were still in development, even during full-rate manufacturing. It also frees every engineering team to iterate as often and as fast as required on their respective part of a product, regardless of what other teams are doing on other parts of the product. Not only does this make our products better and keep them at the leading edge of technology, but it also allows us to wring complexity constantly out of those products and make them easier, faster, and cheaper to manufacture at higher rates. It is simply impossible to achieve the exponential increases in defense production capacity that we need any other way.