How to Make Hardware Venture-Shaped

For the last few months, every conversation I have with another investor eventually drifts to the same place: are you guys doing hardware? Always a little sheepish, like they're confessing to something. We're thinking about it. We know we're not really supposed to.

They're all doing it. American Dynamism is back, reshoring is real, AI has turned energy and compute into physical bottlenecks. Investing exclusively in software is suddenly a limitation instead of a discipline. The whisper network has decided that atoms are cool again.

Here’s the thing: the reason to be excited about hardware is also the reason most of it will lose you money. While software has become trivially easy to clone, hardware remains appealing because it’s defensible. (You can’t slopship a fuel cell!) The moat is real—but in hardware, moat was never the hard part. Of course it’s hard to build. That’s the whole point. 

But that appeal is a trap, because hardware companies almost never die from a lack of defensibility. They die from the financial physics of capital intensity, one-time sales, and dilution. Most hardware companies fail that test long before investors ever get a look.

So the useful question isn't “should we do hardware.” It's “which hardware is actually venture-shaped?” When I run our recent deal flow through that filter (lately it’s been a lot of magnets, PCBs, nuclear, and robotics in my inbox) the ones that clear it all share a structure. As far as I can tell, there are four ways to give atoms a venture shape:

• Razor blades. Hardware is the wedge into a recurring consumable, service, or data layer. You earn the box once and the margin forever.
• Chokepoint supplier. The company becomes one of very few qualified providers of a scarce input downstream buyers urgently need. Boring but scarce.
• Software wrapped around atoms. The physical product creates the footprint; software and data drive the margin and the retention.
• Outcome seller. The company sells the result (uptime, completed jobs, power delivered) not the box itself.

Let's take them one at a time.

Razor blades

The unit is the wedge; the business is what the unit consumes. Printers were never about printers. Instruments make their money on reagents, machines on service contracts and uptime, sensors on the data they generate forever after they're installed. A one-time hardware sale re-earns its entire cost of goods on every unit and caps out at a product-sales multiple. A razor-blade business earns the box once and the margin recurringly.

If the answer to "how do you make money" is "we sell the unit at a good margin," you don't have a razor-blade business, you have a manufacturer. And manufacturers don't return funds.

Chokepoint supplier

You become one of very few qualified suppliers of something everyone downstream urgently needs. This is the picks-and-shovels position, and it's the strongest one on the list. Rare-earth magnets, HTS wire, specialty substrates, defense-grade components—these are the boring but scarce inputs that nobody notices until they're the reason a $100B buildout can't ship.

The moat here is qualification. Once you're designed into a customer's system and certified, ripping you out means re-certifying, which means twelve to twenty-four months nobody wants to spend. Qualification gives you multiyear lockin.

The failure mode to watch in this category: confusing scarce right now with structurally scarce. Timing scarcity doesn’t last; the moment prices get attractive enough, competitors will build capacity and the moat evaporates. A real chokepoint doesn't work that way, and qualification is usually what makes the difference.

Software wrapped around atoms

The box gets you in the door and creates the footprint; the margin and the retention come from the orchestration, monitoring, workflow, and system-of-record value layered on top. The hardware is the beachhead, not the business.

This is the cleanest of the four because it drags the economics back toward software. Think of the AI vision and application layer riding on top of industrial robots that the OEMs will never build themselves: the physical thing earns the right to sell the software, but the software is what makes it venture-scale.

Outcome seller

The company doesn't sell equipment at all; it owns a measurable job to be done. Customers don't buy the machine; they buy uptime, completed jobs, power delivered, deviations resolved, admin burden removed. This is where our services-as-software thinking lives, the same instinct behind backing companies that wrap software around a real-world operation and get paid for the result instead of the tool.

It's a razor-blade taken to its logical end: you're selling the shave, not the blade. When it works, it's the highest-retention shape of all, because the customer has outsourced an outcome they can no longer imagine doing themselves.

And then there’s the law of physics

Even the right shape can’t survive the wrong capital stack. Venture equity is the most expensive money in the world, and hardware founders keep using it to buy the single cheapest-to-debt-finance thing in the world—a factory—and then get washed out over a decade of dilution while their seed investors watch 10% become 2%.

The companies that avoid this all have the same discipline: someone other than your equity funds the atoms. Customer-funded pilots, offtake agreements, project finance, government contracts, grants, strategic capital, debt against a purchase order, all great. That money funds the plant, and venture equity funds the company (the IP, the team, the first reference unit).

The best signal I saw in a manufacturing deck this year wasn't the technology. It was a line saying the customer was funding the pilot and paying for certification. That can be the difference between a fund-returner and a science project.

A few things this framework is not saying

First, these aren't mutually exclusive. The best ones stack. Outcome-seller and razor-blade are cousins. Software-wrapped and outcome-seller often go hand-in-hand. The strongest hardware companies I've seen combine two or three of the four, and every single one of them runs capex-light underneath. If a company can only claim one shape and can't tell you who funds the steel, be careful.

Second, this is about shape, not sector. I got here by trying to articulate why a genuinely good, honest, in-thesis fuel-cell company was still a pass for us. Right sector, honest team, real moat. What it lacked was any of the four shapes. That's the most common way hardware dies in our pipeline, and the hardest failure to spot if you're only looking at the technology.

Third, it collapses to two questions. For any hardware pitch, the two questions that do the most work are: what's the recurring or chokepoint layer, and who funds the capex? If a founder can't answer both crisply, you already know the shape—and it's usually a beautiful, defensible business that isn’t venture-shaped.