New Biz

Why is HardTech so Effing Hard?. Are you prepared to build a factory… | by DC Palter | Aug, 2022

Are you prepared to build a factory? Few people are.

A Rice-Based Ethanol Production Line Using Multiple Parallel Fermentation Technique. Photo by Author.

Take a glance at the 1000+ unicorns listed on Wikipedia.

What industries are they in?

The vast majority are software, SaaS, fintech, e-commerce, AI/ML, edtech, or cybersecurity. Software, software, and more software.

Then there are a bunch more in food delivery and health platforms. In other words, more software.

And then comes all the gaming companies and crypto/blockchain/NFT startups—yet more software.

Outside of software, the only unicorns building actual physical products were 2 in space, 1 in batteries, 2 in biotech, 1 in nanotech, 4 in robotics, and 4 in semiconductors. I’m sure I missed a few, but not many.

That means less than 2% of successful startups actually build things, and most of those are electronics that are easy to outsource.

Where are all the startups in climatetech, chemicals, and new materials?

Where are the new batteries, green hydrogen, sustainable materials, plastics recycling, and more efficient solar panels?

Where are all the carbon sequestration startups, locking CO2 in the atmosphere into useful products? Where are all the startups making steel and concrete produced without coal, plastics without oil, and sunscreens that are safe to use? Where are the next generation of Apples, Intels, Hewlett-Packards, and the companies that built Silicon Valley?

They’re not on the list of unicorns.

Do they not exist?

I know for a fact they exist. I see dozens of hardtech startups every week. I invest in them, I mentor them, I attend their conferences and participate in their Slack boards. I’ve even built two successful electronics startups myself.

They exist. But none have become unicorns. They’re not valued highly. Very few survive.

If there’s one overriding reason why almost no hardtech startups have become unicorns, it’s this: building a hardtech startup is really effing hard.

Software is a venture capitalist’s dream. Write a check, make a couple of introductions, and come back in 5 years for the party when the company has a huge exit.

A handful of software developers can build a prototype in their spare time. Almost everything they need is free. If any specialized skills are required, they can usually be found on Upwork.

Even before fundraising, a software startup usually has an MVP, sometimes even paying customers. Add a few hundred thousand dollars to begin marketing. Then add a couple of million to scale. Then a few million more to take over the industry. Easy-peasy, capital requirements measly — zero to the moon in five years.

Software is infinitely scalable. Just hook up AWS and go from ten users to ten million overnight. No mess, no fuss, just add more sales guys and a couple of support people overseas.

The skill sets are widely available. A python programmer is a python programmer whether they’re working on a fintech app or e-commerce. The team can work from home in San Jose, Costa Rica, as easily as a WeWork in San Jose, California.

When things aren’t working, it’s easy to find answers on Stack Overflow. Worst case, you can switch to a different tool or download the source code and fix the problem yourself. Try that when you’re developing a new Li-ion cathode.

Want to make a better battery electrolyte or invent a way to recycle plastic? You’ll need all kinds of specialists, you’ll need a lab, you’ll need capital for inventory, and you’ll need to ship a product. You’ll need to build a friggin factory. A FACTORY!!! Ever done that before? Me neither.

I don’t want to say a software startup is easy. It isn’t. 90% fail. But a hardtech startup is 100x harder. If software is a venture capitalist’s dream, hardtech is her nightmare. It takes specialized skills. It takes millions in capital just to find out if the technology is viable. And it takes investors who understand technology instead of spreadsheets.

Most of my career has been spent at two computer networking startups. As much as we wanted to be software companies, and as much as we were fundamentally software developers, our product was a piece of hardware that customers installed on their networks.

I thought it would be easy. We didn’t have to build a factory ourselves. There were plenty of electronics contract manufacturers to manufacture the product for us, from the small aerospace shops down the street to the city-sized factories of Flextronics and Foxconn.

The devices we needed used off-the-shelf components with some minor modifications. How much easier could it get? With the hardware taken care of by professionals, could I spend my day on corporate strategy? Talking with customers? Directing the marketing?

No, I was dealing with hardware problems. Begging the manufacturers to build the hardware we needed at volumes that were in the noise for people used to working with Apple and Cisco. Finding space in our office and cash in the bank account for a year’s worth of inventory because that’s the minimum quantity they’d manufacture for us.

When the inventory arrived and was stacked six feet high against every wall (don’t tell OSHA), I thought my life would be easier. Now I could focus on selling the boxes to customers.

Wrong. Pretty soon, I heard from customers that some of the units wouldn’t boot up. It turned out the manufacturer used a batch of expired motherboard batteries. Some died in days; some worked for years. But I had to organize a recall of hardware installed in customer networks worldwide. Just kill me.

Even when things were going well, I spent much of my day apologizing to customers when FedEx lost their shipments, arranging for FCC, CE, and UL regulatory certifications, then figuring out how to get the same approvals for Japan, Australia, and a dozen other countries when our shipments got locked in customs.

I once spent an entire Christmas season in Taipei doing everything I could to get the electronics manufacture to figure out what was wrong with all the boxes they’d shipped us that were failing at customer sites. It turned out to be a little detail, like they were using an oscillator that was intended for PCIe v2 on a PCIe v3 network card. It worked most of the time, but not all of the time. I lost months of shipments and 75% of my stomach lining over the wrong, stupid $1 chip. The only thing that saved me from a hospital bed was prescription-strength Prilosec.

My worst experience was contracting with a local electronics firm for a specialized product the big guys weren’t interested in building for us. That factory declared bankruptcy the week we started shipping our hardware. We needed a court order to get the sheriff’s office to open their gate and let us take our half-assembled hardware back to our office to finish building ourselves.

Can you say aargh? Because nearly every day, I was screaming about something. From the sales tax audits to broken hardware, to lost shipments, to having no inventory at all. Every hardtech CEO needs a private office, not for quiet contemplation, but for screaming really loud.

But if this sounds bad, my complications were minor. This was basic electronics manufacturing with an outsourced contact manufacturer. The problems were minuscule compared to real hardtech — new chemistry, materials, and hardware systems.

I work with many early-stage hardtech startups, many of which were founded by grad students who developed interesting technology in an academic environment and are looking to commercialize their inventions.

The usual answer to the capital cost and complications of building a large-scale production and sales operation is to license the technology to potential customers.

It sounds easy. It sounds fun. 3M or BASF will pay 10% of their product cost and make and sell the product themselves. The founders can sit back and focus on R&D, which is what they really want to do, while the cash rolls in.

It’s a great idea—just 2 little problems. First, 3M or BASF is unlikely to license your technology instead of developing their own, and certainly not anywhere close to 10% (1–3% is typical for licensing deals).

Second, licensing is a great business model for bootstrapping a business, but is not suitable for venture capital investment — licensing businesses can be highly profitable, but they don’t generate big exits. The only exit is usually private equity which will only pay a small multiple of EBITDA.

So, if you’ve invented a great technology, you’ll have to figure out how to scale it up yourself. You’ll have to figure it how to manufacture it. You’ll have to deal with supply chains, factory workers, shipping, and selling to industrial customers. If you want to focus on research, you’re in the wrong business — QC is a more important function than R&D, inspectors more valuable than scientists.

There are toll manufacturers who can build and run a factory for you. They can handle some of the logistics. That can help as you’re getting started. But it’s your product not theirs, and you need to treat it as your factory, not theirs, as a rental facility rather than outsourcing production.

If you want to build a hardtech startup, be prepared to think through how to deal with these challenges from the start:

  1. The capital requirements to build a production facility and maintain inventory are orders of magnitude larger than writing software.
  2. The time required to scale from prototype to test to production are multiples of getting a software product out the door.
  3. A software bug can be fixed with a patch in a few days. Even a small chemical production problem is a disaster.
  4. Founders are usually scientists with little business skills. Even with a business person as co-founder, that usually means sales and marketing, not building and operating a factory and dealing with supply chain issues.
  5. Acquisitions tend to be based on a small multiple of EBITDA rather than a huge multiple of revenues.

In other words, hardtech costs more, takes longer, has a higher risk of failure, and pays out less on success. No wonder the VCs stay away.

Put these together, and you have a difficult challenge — how to get from an important invention into profitable production and, eventually, a successful exit.

With all these complications, is there any reason to build a hardtech startup and invest in hardtech companies? Yes! These are the innovations we need. Urgently. To solve our climate crisis. To build a better world.

Software is great, but there are only so many new social media platforms we can waste our time on, and there is a limit to the number of meal delivery apps that will improve our lives. And don’t get me started on blockchain. If we devoted even a fraction of the time, skills, and electricity to solving energy storage as we do on building more cryptocurrencies, the world would be carbon neutral by now.

Building a hardtech startup is nearly impossible. But I beg you to do it anyway. Because the world needs real physical solutions to hard problems far more than another piece of software.