Key Takeaways

  • Pure software APIs fail in hardware discovery because semiconductor fabs protect trade secrets behind air-gapped security perimeters.
  • Periodic Labs seeds forward deployed engineers onsite to run inference locally and fine-tune foundation models on proprietary fab data.
  • Liam Fedus plans to transition commercial pricing from software copilot seats to pricing guaranteed synthesis outcomes as machine autonomy increases.
  • Ekin Dogus Cubuk argues commercial profits in solid-state physics will reverse the talent drain from physical sciences back to hardware labs, echoing Bell Labs.

The Air-Gapped Reality of Materials Science

Silicon fabs do not connect their manufacturing data to external web servers. If you build an AI model for materials discovery and expect semiconductor companies to send proprietary chemical formulas over an API endpoint, your sales pipeline will stall out.

Liam Fedus and Ekin Dogus Cubuk recognized this roadblock early at Periodic Labs. Instead of pitching a remote software subscription, Periodic borrows a tactic from Palantir: forward deployed engineering. Their staff work directly inside the client facility. Fedus explains the strategy:

“We've been building our own products, our own tools, our own AI and computational for ourselves. So we've been customer zero. Now what we're doing is we're taking these same tools to different industries and we have a huge focus right now in the semiconductor industry. Our periodic forward deployed engineers and researchers will actually be onsite working and actually integrating our AI systems or computational systems to help our partners get to their goal states more quickly.”

These teams do not simply write scripts. They understand solid-state physics and run inference locally on internal servers. Fedus points out the compounding advantage: “Then these forward deployed engineers will also do the inference locally but then also can train on the data. So again once we take our system that understands these different areas and we deploy we can make it expert on the customers or the partners' data so that they can own their own intelligence and have systems that understand this much more effectively than just hitting some API or some untrained model.”

From Assistant Copilot to Pricing Outcomes

Commercializing high-risk science requires meeting customers at their current comfort level before charging for autonomous work. Early software tools like GitHub Copilot did not write full codebases automatically; they shaved minutes off routine functions. Over time, that software grew capable enough to run autonomous workflows.

Fedus maps Periodic's commercial roadmap directly to that software trajectory:

"In the early days of software engineering like we had GitHub Copilot, then we had early versions of ChatGPT. People were using these things as copilots to help them get to their solutions, and as the automation improved now we have things like Codex and very few of our engineers are writing code the way they used to. And we think a very similar thing could play out for periodic as well where you have systems to accelerate the researchers, the material scientists, materials engineers, the process engineers. But as the autonomy, intelligence, and capability grows, you can begin to price outcomes."

When a model acts as an assistant, you charge for seats. When the model discovers a stable thin film or solves a yield failure independently, you charge for the result itself. Moving up that pricing ladder requires trust that can only be built while sitting next to the plant operators.

Rebuilding Bell Labs Economics

For thirty years, digital software captured the bulk of venture capital and elite engineering graduates. Writing social apps or web services offered rapid liquidity, while materials research meant grinding through decades of slow academic tenure tracks. Cubuk wants commercial materials research to change that incentives balance:

“Our best contribution to solid state physics and science could be if we made these tools and topics profitable. Similar to how ChatGPT made CS and LLM majors way more popular in colleges before and after. We'd love to show that solid state physics material science research can make a big impact commercially and then that will attract more attention.”

Bell Labs produced the transistor and the laser because commercial profits paid for basic physics research. By converting deep tech into immediate enterprise revenue, Periodic wants to build a similar talent magnet for the physical sciences.

What to Do With This

Audit your go-to-market plan for any customer who handles regulated or top-secret data. If your current sales cycle relies on convincing enterprise clients to send sensitive inputs to your cloud endpoint, kill that requirement this week. Package your model for a local test instance and offer to embed one of your engineers inside their office for a two-week proof of concept.