美国制造业资产类别的投资逻辑

Today, we launched Discourse, a curated forum on top of the Cosign network. 3,000 people have access to post directly, and everyone else can submit posts into a queue, kind of like old Product Hunt. Check it out: https://discourse.cosign.co/

America’s defense-tech renaissance is well underway. If you’ve spent time in the Gundo, Silicon Valley, or even Washington, D.C., that should not surprise you.

Venture dollars are rolling into defense in ways nearly no one thought possible, funding companies like SpaceX, Anduril, Castelion, Saronic, Covenant, and K2 Space that have turned once-improbable ideas into advanced missiles, autonomous systems, aircraft, ships, and spacecraft. They move rapidly from design to prototype and often deliver cheaper, highly performant systems in place of one-off exquisite platforms.

Yet raising capital and winning development contracts aren’t what ultimately matter for great defense companies. What matters is production and fielding at scale. Without it, America ends up with prototypes instead of an overwhelming force capable of deterrence. That push to scale is now sending a surge of demand through an aging industrial base that can’t keep pace. Silicon Valley has largely overlooked that base: we haven’t funded enough of the manufacturers and supply chain beneath these companies.

That gap is a national security liability. It is also a generational opportunity for American capital to build enduring manufacturing companies capable of generating meaningful returns, while unlocking high-skilled jobs and industrial capabilities with positive ripple effects throughout the broader economy.

This opportunity is arriving at the exact moment the United States has made a top-down political decision to reduce its reliance on China. Batteries, motors, actuators, turbines, rare earths, and other parts of the industrial stack cannot remain exposed to Chinese leverage.

America cannot compete on labor cost, nor should we try. Instead, we need to use our greatest asymmetric advantage, deep capital markets and the technological innovation they fuel, and engineer our way out.

That means building a technology-first industrial hierarchy from the ground up. That hierarchy runs from Boeing at the top, to Honeywell beneath it, down to the machine shops several tiers below that make precision components. The goal is an industrial base every bit as modern as the advanced systems built on top of it.

Modern Defense Tech Has Outrun Its Production Base

Before re-engineering the manufacturing base, it helps to understand how modern defense companies build systems. When a new advanced system flies, sails, or drives for the first time- take Castelion’s Blackbeard, one could reasonably ask: you built a handful. Can you do the same thing 10,000 more times?

Unsurprisingly, the unit economics that get one missile through a flight test are nothing like the economics of a production line running at scale. To finish a prototype, a company may pay an absurd price for a one-off part, wait on a supplier, or buy tooling that never pays for itself. These are rational decisions when the priority is proving something genuinely new and transformative for the Department of War.

However, once companies want to move a product into large-scale production, the challenge shifts from proving the capability to building scalable manufacturing across the existing industrial base. Castelion is working their Blackbeard hypersonic missile through that transition now through a recent Navy production order calling for hundreds of missiles annually, with a pathway into the thousands.

As more companies look to make this same transition to scale, that production ramp will need to extend several tiers down the supply chain, into the thousands of machine shops, component manufacturers, and process specialists these systems depend on. The 2022 Economic Census counted 16,876 machine shops; among those operating all year, 83% employed fewer than 20 people and 95% fewer than 50. The same small-firm structure extends across American manufacturing, where roughly three-quarters of the country’s 240,644 manufacturing employers employed fewer than 20 people.

Many of these firms excel at the difficult processes they perform, but companies this small cannot casually add machines, workers, qualification runs, and inventory for an abrupt production ramp. And size isn’t the only constraint on scaling. Fragmentation across the industrial base is the other. The capabilities needed to build a single system are scattered across highly specialized shops that rarely function as a coordinated network. In many critical processes, there may only be one or two qualified sources to begin with. Strong industrial clusters have historically produced deeper pools of specialized suppliers, workers, and technical knowledge. Without a robust industrial network, one supplier several tiers down may impose a production ceiling on the entire end system.

These tensions exist because much of America’s supplier base was built around low-rate production of exquisite systems with service lives measured in decades. Suppliers were rewarded for doing specialized work in small runs, not building excess capacity for systems ordered by the thousands. 61% of tier-two-and-below defense manufacturers identified tooling, automation, or production-line limitations among their top three barriers to expansion. And that work cannot simply move to whatever factory has room, because qualification can be tied to specific manufacturing processes and facilities.

So the challenge is not as simple as just adding more capacity. The high-low mix our nation now requires asks the existing supplier base to sustain exquisite systems while ramping entirely new classes of low-cost, high-volume weapons and systems. Doing that requires more qualified production across the tiers, but also a supplier base that can actually work together as one industrial system. We need to modernize both the factories and the engineering relationships between them.

Not only does this require enormous upfront capital, but especially in defense, systems companies and their suppliers are often expected to prepare for a production ramp while the government remains noncommittal on how much it will buy or when.

Factories Ramp on Orders, Not Maybes

The supplier gap ultimately comes down to two constraints: uncertain defense demand makes the next production ramp hard to underwrite, while gaps in technology and manufacturing capability limit what suppliers can actually deliver at scale.

Private capital can build these businesses, but it cannot rationally finance unlimited defense-specific capacity against demand the government has not committed to. Most machine shops were built to quote drawings and fill purchase orders, not carry the machines, workers, inventory, and qualification costs of a speculative ramp indefinitely.

With Anduril’s Ghost-X program, suppliers needed roughly six months to add capacity, but the Army couldn’t commit to how many systems it would buy, forcing Anduril and its suppliers to weigh scaling costs against uncertain demand. And this is now extending into the multibillion-dollar CCA program, where Anduril says it has reached the limit of what private capital can carry and cannot begin FQ-44 Fury production without the Air Force’s $1.1 billion FY27 procurement request being funded. Sellers need buyer dollars. Not exactly a crazy proposition.

If we want the most innovative defense companies to build the world’s most advanced arsenal, the Pentagon has to act as the anchor customer. Those commitments do two things: they give investors a credible demand signal to finance new capacity, and they give companies enough certainty to commit to their suppliers before full-rate production begins.

Reliable demand, however, only unlocks the capacity decision. Turning that demand into thousands of systems still depends on a technology-first supplier base with the engineering depth and production capability to deliver at scale.

Build-to-Print Is Not Enough, the “Demand and Supplier” Capability of the Future Must be Co-Engineered Together

One of the core defects in today’s supplier relationships is a missing engineering feedback loop between systems companies and component manufacturers. Companies that master this will define the next century of manufacturing. Hint: private capital, invest here.

Traditionally, a build-to-print vendor gets a drawing, quotes it, and makes the part. Whether that part is easy to make or will produce scrap at volume is not the vendor’s call. Suppliers often know exactly why a part is costly or difficult, but they have no authority to redesign it and no reason to try. The result: production knowledge never flows back into design, and prototype decisions get frozen into programs meant to produce tens of thousands of units.

The more modern approach starts with the performance requirement and system constraints, not the inherited drawing. The supplier should be able to tell a high-performance missile company:

“We can build the fin this way. It will still meet the performance requirement. Oh, and by the way, we can produce 20,000 of them. Way cheaper, too!

That is co-engineering at its core: redesigning the component and its production method together, questioning the materials, geometry, tolerances, interfaces, part count, testing, qualification, and even the production process itself. The systems company can still retain its architecture and final integration while the supplier takes bounded responsibility for the component and its production system.

Ultimately, this is a first-principles exercise. It’s also hard, and it requires doing real work from scratch. Using 3D printing or another new method to make the same old part faster is helpful. But asking why the part needs to exist in that form at all is even better.

As enlightening as this may sound, this is not a new concept at all. In fact, we’ve known this for decades. A study of the global auto industry in the ‘80s found that intensive supplier participation in engineering helped explain substantial advantages in development time and cost.

Continuing this collaboration throughout the product development lifecycle will expose intricate and possibly overlooked nuances that could affect end performance or manufacturability. A tolerance that looked harmless on paper may drive scrap at volume, or a design that worked perfectly in prototype may be painful to repeat 10,000 times. That production learning needs to flow straight back into the next design decision.

Running that loop across thousands of components and a broad supplier base requires software that keeps requirements, design, test, and production data connected. Companies like Nominal connect data streams like those and cut Anduril’s five-to-six-hour test-review cycles to near real time.

Master first-principles co-engineering, and every component becomes another source of process knowledge. Capture that learning in software, put it back to work across factory floors, and the supplier becomes critical infrastructure for the next generation of physical systems.

The Whitespace Beneath the Primes

Elon famously said, “The Factory is the Product.” And nowhere was this idea more apparent than at SpaceX.

From the early days, Elon decided to build “almost all” Falcon 9 and Dragon internally. More than 600 Falcon 9 flights later, they’ve extended that logic to their advanced behemoth, Starship, manufacturing roughly 80% of it in-house.

For the companies solving truly once-in-a-generation challenges, verticalizing may be necessary to keep design, manufacturing, and testing in one development loop. Why? Because co-engineering across an entire sub-tier supply base would require too many leaps of faith while the technologies are still being perfected (essentially building the plane in flight). However, most companies are not SpaceX and shouldn’t organize themselves as though they are.

Most systems companies should own what actually defines the product and bring in-house whatever the market simply cannot provide at the required performance, cost, volume, or speed. Beyond that, rebuilding the industrial base inside every new prime makes little sense. At some point, every systems company will need a broader manufacturing base beneath it.

That is the whitespace. That’s the investment opportunity.

Even Anduril, which brought rocket motor manufacturing in-house, still relies heavily on the manufacturing base beneath it. Its low-cost Barracuda-500M uses 70% commodity components, with the remaining 30% de-risked through competing open-architecture designs from multiple vendors. And Castelion is pushing even further into the commercial manufacturing base, designing its low-cost missile seeker around automotive electronics suppliers and standard industry sizes.

The point is not that these companies should outsource everything. They shouldn’t. The opportunity is to build and invest in the suppliers that can own the components and manufacturing processes the primes do not need to own themselves.

And importantly, the economics of that supplier base improve when the same capability can serve multiple programs. When aircraft programs buy common avionics together, they can combine demand into larger orders and capture economies of scale⁠. Proprietary architectures do the opposite, locking programs into unique components that are harder to source competitively.

Manufacturers build more resilient businesses when they don’t have to bet their factory on only one program. The engineering knowledge, tooling, qualification, and production capability it builds for one customer can carry into the next while derisking the supplier long term.

But the broader industrial network needs redundancy too. If one fantastic manufacturer ends up supplying the same critical component across five missile programs, that supplier may have a well-diversified business, but it also introduces a five-program single point of failure. The goal is resilient manufacturing capacity without recreating sole-source dependence.

This is the industrial whitespace Silicon Valley has largely ignored. We funded the companies building the systems. Now we need to fund the manufacturing layers beneath them so every new prime does not have to rebuild American industry inside its own walls.

Capital In, Capability Out

Manufacturing is capital-in, capability-out. More output requires more of everything: engineering, qualification, machines, facilities, inventory, skilled workers, and most of all significant capital. There is no ‘silver bullet’ software-style zero-marginal-cost escape hatch. Consequently, this is one of the few venture and private equity categories that isn’t obvious yet. If America wants more industrial capacity, somebody needs to finance the physical capability.

Capital historically chases the systems companies with the headline contracts. Tier 2 and 3 suppliers, the layer technology hasn’t touched, sit grossly underpriced.

Thousands of qualified suppliers hold assets built over decades: approvals and qualification history, components already flying on active platforms, skilled workers, deep customer relationships, and hard-won process knowledge. Most are operating at a fraction of their potential because of dated software, engineering, workflows, and equipment. Advanced technology adoption remains concentrated among the largest firms; the long tail has barely been touched.

However, let’s be clear: the opportunity is not to buy these shops and slap AI on them. It is to use modern engineering, software, and capital to make scarce industrial assets produce more.

This opportunity exists partly because of the tailwinds behind the intelligence and technology that connect the entire manufacturing process. Advances in smart sensors, industrial IoT, AI, and modeling and simulation now give manufacturers a much clearer view of what is actually happening inside a production process, while Physics-informed AI and digital twins let engineers use production and inspection data to improve the next design and process. That same visibility can extend across suppliers, exposing capacity and bottlenecks before one shop becomes the production ceiling for the entire system.

But adding software, no matter how advanced, doesn’t change the basic fact that manufacturing requires machines to make real, tangible products in the physical world. Every dollar of technology still has to contend with whatever is physically limiting the line.

Enter the two-second transfer test. If a worker can move a part between two machines in two seconds, adding a robot probably adds cost without meaningfully increasing output. Manufacturers need to ask what is actually constraining production. Is it the cutting cycle on a CNC mill, the time spent changing fixtures, or parts sitting around waiting for inspection? That is where technology and capital should go to work.

Two proof points putting technology and capital to work on factory floors, both from differing directions:

Hadrian builds new, AI-driven factories from scratch. Its Opus platform turns customer design files into manufacturing steps and work instructions, schedules machines and technicians, and monitors production through inspection. Even in 2022, Hadrian was already making spaceflight-grade parts 10x faster and more than 40% more efficiently than the legacy supply chain. Then, on RTX programs, Hadrian-made Javelin and TOW components achieved 98% on-time delivery.

Amca combines AI-based engineering with established manufacturing capacity to deliver new critical components. Why spend years recreating customer approvals, qualification data, and process knowledge that already exist? Its RAPID engineering platform digitizes decades of tribal knowledge and pairs it with custom simulation tools and in-house prototyping and qualification to generate and validate new, manufacturable component designs. Across their six factories, Amca ships more than 50,000 components each month and, using RAPID, has cut development-to-production timelines by 67% against industry standards.

One builds new capacity from a blank floor plan. The other modernizes and grows the capacity America already has. Both show how technology and capital can expand the supplier layer beneath the primes instead of forcing every systems company to build it themselves.

That is what makes this an investment opportunity, not just an industrial-policy problem.

So with this modern, technology-first approach, capital in = capability out, right?

Yes, and now more than ever, that capability can be exponentiated.

One of America’s greatest strategic advantages is its uniquely deep capital markets, capable of financing reindustrialization from the first defense production line through repeatable factory expansion and into much larger commercial markets. Government capital must act as an anchor and fill the gap where a capability is strategically necessary because private capital cannot always carry it alone. Venture can then take the early technical and operating-model risk.

Once production is proven, growth equity, private equity, strategic capital, and credit can finance the next factory or stage of production. At some point, expensive venture equity becomes the wrong tool for expansion. Where a manufactured product has broader use, allied and commercial demand can sustain capacity utilization and support the next expansion. That is America’s capital markets working in harmony with manufacturing to strengthen the broader industrial base that defense ultimately depends on.

The Defense-Industrial Flywheel

Hadrian’s progression from venture-backed factories to larger equity and factory-expansion financing shows this handoff in action. One week after raising $1.37 billion in equity, Hadrian closed a $360 million revolving credit facility specifically for manufacturing infrastructure, machinery, and hardware. Then institutional entrants like Apollo and JP Morgan signal that industrial production can attract larger, cheaper pools of capital. So, contrary to instinct in most venture-backed markets, private equity flowing into American manufacturing can be a good signal.

Critics will say this is exactly how private equity hollows out American manufacturing. Fair enough. The standard should be simple. Does the capital leave behind stronger engineering capability, qualification capacity, equipment, skilled workers, reliable output, and independent sources? If yes, great! If the plan is to buy an aging shop, lever it with debt, and hollow out its productive capacity, that is not industrial renewal. It is stripping the industrial base for cash. Acquisitions should leave America with stronger suppliers, not fewer qualified sources. The goal is more winners, not fewer owners. Anything otherwise is dangerous to the national interests.

Deterrence Starts on the Factory Floor

To prevail in any large-scale conflict, America must build a durable industrial base capable of turning the past decade’s defense-tech renaissance into an overwhelming arsenal - one that makes clear a war with the United States is not a war worth starting.

To do this, we must rebuild our industrial base technology-first. We will not debase our Western ethics to chase Chinese labor costs, nor imitate a model that subsidizes uneconomic overcapacity to dominate critical supply chains and weaponize that dependence in a crisis.

We will win differently, by leveraging frontier technology and the world’s deepest capital markets to build productive capacity that is economically durable on its own.

That makes this industrial build-out a national security imperative and, at the same time, a massive opportunity for American capital. The two aren’t mutually exclusive.

We are entering a new age of manufacturing, one that may be inspired by Silicon Valley, but will be built across Middle America, creating what will become a brand new American manufacturing asset class.

Subscribe now

This newsletter is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. Furthermore, this content is not investment advice, nor is it intended for use by any investors or prospective investors in any a16z funds. This newsletter may link to other websites or contain other information obtained from third-party sources - a16z has not independently verified nor makes any representations about the current or enduring accuracy of such information. If this content includes third-party advertisements, a16z has not reviewed such advertisements and does not endorse any advertising content or related companies contained therein. Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z; visit https://a16z.com/investment-list/ for a full list of investments. Other important information can be found at a16z.com/disclosures. You’re receiving this newsletter since you opted in earlier; if you would like to opt out of future newsletters you may unsubscribe immediately.