Cameron Porter is co-founder and general partner at Steel Atlas, a VC fund investing in the software and hardware that is re-shaping the industrial economy.
Tell us about yourself and Steel Atlas.
I’m a co-founder and general partner at Steel Atlas. We’re a firm based in New York City investing in software and hardware technologies re-shaping the industrial economy. Before Steel Atlas, I worked in venture for Kevin Ryan at AlleyCorp, which has incubated many of New York’s biggest companies (MongoDB, Business Insider, etc). Prior to that, I was a pro-athlete in Major League Soccer.
At Steel Atlas, we view technologies through the lens of resilience and volatility. We look at the inputs to industrial assets like power prices, mineral prices, and freight prices, as well as factors that create operational uncertainty like downtime, production bottlenecks, labor shortages, and construction delays. We then ask a simple question: What technologies can mitigate or eliminate these forms of uncertainty? It’s similar to how industrial asset owners and operators make decisions about the technologies they adopt.
A weekly dispatch featuring exclusive interviews with deep tech founders & a roundup of the most important deep tech news.
What types of companies do you normally invest in?
We spend a lot of time focused on the supply chain, given its importance to the industrial economy. For example, we recently looked at freight price volatility and discovered that a major issue is market liquidity. Despite GPS and other tools that should allow tracking of long-haul trucks in the US to match supply with demand, it’s all done manually by about 27,000 brokers. These brokers have limited knowledge of available trucks, and so it’s difficult for the market to be efficient.
The obvious solution would be tracking all trucks, but it’s hard to convince more conservative operators to put tracking devices on their vehicles. The result is that only about 10% of the market is tracked at best. We led the seed round in a company called GenLogs that was started by someone who had spent 15 years in the Middle East tracking terrorists using public camera infrastructure. His company has now built the largest sensor network on the US highway system. We use these signals to de-anonymize other data sets, producing a real-time tracking system that now monitors the majority of U.S. freight. This is emblematic of the type of businesses we like to support. Even though they might be offering hardware upfront, once we fund and build out their networks, they end up looking like a software business.
How does this approach apply to other investments?
The core archetype for any great venture company is that you want a fixed amount of dollars going in and a large, defensible cash flow coming out. We even see this pattern in areas people might not think of as classic venture businesses.
Most nuclear companies, for example, don’t own and operate reactors. They spend a fixed amount of capital designing them and then license those designs to operators in different countries. If you’re willing to spend the upfront cost to get something licensed, you can produce a high-margin, cash-flowing product where you can receive licensing fees for a minimum of 40 years, which is the operational lifetime of a reactor. The beauty of industrial investments is that they touch things people think are difficult. As such, they’re much more defensible, especially in the age of agentic AI systems further lowering the barrier to the production of purely digital goods and services.
What’s your perspective on public-private partnerships in deep tech?
They’re incredibly important. They serve a key role in bridging the “valley of death” for deep tech or hard tech businesses, especially with large first-of-a-kind facilities. Startups may have a great idea at pilot scale, but they’ve never delivered it at industrial scale. They need someone to take the risk to help them build that first facility. Eventually, technologies get to the “nth of a kind” and use private markets for equity and debt to scale. But the question is how to bridge that gap in the middle.
At Steel Atlas, we’ve spent time in different global markets to understand where things work well and where they don’t. Saudi Arabia provides an interesting contrast to the US. They’re undertaking a massive diversification effort with Saudi Vision 2030 to move their economy beyond oil.
One thing that works surprisingly well there is their focus on demand-side support rather than supply-side. Instead of just providing financing or tax credits like the IRA, they’ll say, “We need to buy X amount of high-tensile strength steel from a local supplier at this price.” They don’t dictate how to do it; they just create guaranteed demand. With that revenue guarantee, companies can secure private financing to build facilities and adopt new technologies.
What do you make of the US’ current approach to these partnerships?
When I look at the US, things tend to go well when we focus more on inducing demand versus providing supply support. Intel is a perfect case study. We really want domestic chip manufacturing, but just making it cheaper for Intel to build facilities doesn’t ensure they’ll build viable ones. Intel needs demand. If government agencies like NASA or DOD committed to buying specific quantities of chips with certain performance specs, that would be much more effective at driving competitiveness and productivity.
On the demand-side partnership front, I don’t think we’ve clearly gotten better with the new administration. We’re probably on par, with certain areas better and others worse. Where I’m more optimistic is on the regulation side. Free market movement is going to make the US more competitive. Entrepreneurs need to have a well-oriented compass with minimal distortions, that lets them pursue good opportunities quickly. That’s how the US can remain competitive against more centralized players like China.
It’s similar to what I think we get wrong with AI policy. Peter Thiel has suggested that crypto and decentralized technologies favor democracy and individual freedom, while AI, because of data requirements and coordination capabilities, would favor centralized regimes. This misses the core value of AI, at least how I use it. AI increases your rate of exploration. It means I can start a company I couldn’t before because I don’t need to hire a mechanical engineer today. I can create a CAD design or write software I couldn’t write before. If we limit Americans’ access to models, including open-source models or models from other countries, that directly constrains our rate of exploration of new ideas.
It reminds me of the Mighty Ducks movie. Americans used to be the creative underdogs who win because we all have different quirky skills and figure out how to use them together. That’s like using AI in cool ways. The Icelandic team (proxy for centralized regimes) in the movie were the bad guys. They cheated, were super strong, and followed their coach’s orders. It turns out that’s not how you win, yet the US is acting more like a centralized regime than the Mighty Ducks. We need to be way more like the Mighty Ducks.
And what about Europe’s approach to public-private partnerships?
CERN is one of the world’s largest research labs and a great example of a public-private partnership. It focuses particularly on particle physics. It wasn’t started specifically just because particle physics is inherently valuable, but because Europe wanted to reverse brain drain. During World War II, scientists were moving to the US, and Europe needed a solution to attract and retain talent. CERN has been quite effective, with a billion-dollar annual budget and tens of thousands of scientists involved.
But, despite all their groundbreaking research, you’ve probably never heard of CERN spin-outs. Tim Berners-Lee developed the Internet at CERN before DARPA took it over. They even developed the touchscreen at CERN before anyone else. But you don’t hear about CERN being the forebearer to the iPhone.
The challenge is that with 30+ countries contributing to the budget, where do you start a company based on CERN research? Do you put it in the UK, Germany, France, or Poland? How do you allocate IP? Does it offset future membership fees or go to scientists’ home countries? These are difficult questions, and the easiest solution has been to prioritize pure science and not commercialize technologies.
This approach worked when Europe was viewed as an extension of the US under the NATO umbrella. But now, as global dynamics shift, CERN can no longer allow technologies to sit idle or let Americans commercialize them. There’s a broader push to rethink how to extract more value from these institutions filled with incredible scientists and engineers. This is happening not just at CERN but at ITER and other institutions as every country reconsiders its role in critical industries and technologies.
The result will be much more competition than before, which I think is positive for the ecosystem. That’s partly why we’re spending more time in Europe. You want to find fertile ground where the flowers are about to bloom. Europe is definitely in that position, with Switzerland as a leading example.
How do you define deep tech?
Deep tech is anything that requires a real portion of technical underwriting to do the deal. There are really two types of risk: market risk and tech risk. Market risk means the technology can be built, but the question is how to get defensible or scalable distribution that lets you outrun competition. Tech risk asks whether something can be built at all, with specific technical requirements. For example, in batteries, what energy density and power production can you achieve, and what is the levelized cost of energy (LCOE)?
Deep tech sits closer to the tech risk side of that pendulum. Most of the underwriting involves figuring out what milestones people can credibly hit in certain timeframes, when they’ll hit commercialization thresholds, and whether you believe there’s a large market once they do. Deep tech should take very limited market risk. If it’s taking both tech and market risk, it’s probably a problem.