What we look for in a semiconductor startup
Most of what we look for in a hard tech company is consistent across sectors: unique IP, quantifiably different value prop, capital efficiency, and a team we believe in. A few months back Andrew Haughian gave a seminar (with our partners at Hakuhodo DY Holdings) to share some insights about our recent Hard Tech Report. After diving deep into the various findings of the report (download here for more detail), Andrew took some time to share some characteristics that semi and AI infrastructure companies should possess that are different from those in other sectors.
Here’s where we see the important distinctions:
Slot Into the Industry Roadmap
A new technology can be impressive on its own terms and still go nowhere if it doesn't solve a critical problem for the industry. Andrew calls what we look for slotting: your technology fits into the industry's roadmaps, answering a step the industry already plans to take. In our view, this is imperative to success in semiconductors and AI infrastructure. If you don’t have a clear and defensible idea of where you slot in, keep at it.
Capture Enough Value
Technology startups naturally aim to deliver a lot to the end customer. The question for the startup is whether it can capture enough of that value for itself. This might be measured in dollars per wafer or per chip: how much your technology moves that number for the customer, and how much of the movement you can charge for.
Collaboration Partners Are Mission Critical
In other sectors we often treat revenue as a key signal when evaluating a potential investment. For semiconductors, ecosystem partners often carry more of the decision. It’s important to have clarity on who a startup is working with and whether they are leaders in the field, because partnerships are how these technologies get to market. Ideally there is more than one.
Pangaea’s newest portfolio company, NLM Photonics, is a useful example. NLM develops Selerion™ electro-optic materials for silicon photonics, aimed at moving data inside data centers with less power. A material only proves itself once it sits inside a working device, so NLM’s partners are how it gets there. Spark Photonics provided design services for NLM’s modulators and chip layouts, including a custom wafer run on AIM Photonics’ 300 mm platform. SilOriX, a spin-out of the Karlsruhe Institute of Technology, brings silicon organic hybrid device design and prior experience with NLM’s materials, and the two signed a joint development agreement in September to expand the use of Selerion-HTX in high-speed optical communications. Tokyo Ohka Kogyo, Hamamatsu Photonics and Mitsubishi Chemical have also invested.
Industry Knowledge
Some of our best portfolio companies were founded by executives who were relatively young and had no direct industry experience. They thought about a particular problem differently and uniquely, and that was part of the key to their success. Andrew's read is that this path is less likely in semiconductors. More likely a team needs to know the industry and has members who worked at some of the larger companies, which often comes with relationships to the partners a startup needs. If a team is early and light on that background, tell us who around you has it.
This list undoubtedly leaves things out, and every investment decision draws on more than these four factors. We see them as the points where semi and AI infrastructure ask the most different questions of a startup, and the likely conditions required for the best chance at taking a technology innovation from a fringe idea to an essential part of the way things are done.