Buyer Chains · CASE STUDY 009 · Deployment at the customer's site
Periodic Labs × an unnamed semiconductor maker: reading what it sells, and to whom, from its job board
The question · When an AI materials company has named no customer, can its hiring tell us who the buyer is and what the buyer wants?
At a glance · The company says it is helping "a semiconductor manufacturer" with heat dissipation on its chips and has named no customer; of 32 openings, one is a simulation engineer stationed at customer sites in Taiwan.
One of the Buyer Chains series. Each piece follows one chain, the AI company, its customer and sometimes the customer's customer, through five roles in the public record: who states the need, who sets the research direction, who runs the experiments, who validates the result, who decides to keep paying. It records only what the public record states and says who said each thing; a missing step is marked not yet public rather than filled in, and a company's own figure is written as the company's. The timeline at the end stays open: when the chain's next step reaches the public record, it goes on as a new line.
Periodic Labs was founded by Liam Fedus and Ekin Dogus Cubuk, who came from research at OpenAI and Google DeepMind; it launched in September 2025 with a $300M founding round, and one of its goals is higher-temperature superconductors (where the two came from and how the company was formed is Lab to Company 006).[1][2] But the first kind of customer it names publicly is not about superconductors: its homepage says it is helping a semiconductor manufacturer "facing issues with heat dissipation on their chips", training custom agents for that company's engineers and researchers to make sense of their experimental data and iterate faster.[1] Its lead investor, a16z, says it is already working with customers in space, defense and semiconductors, with a strategy to "land and expand".[2] No customer, contract or price is public; this piece reads what the company sells, and where, from its own job board.[3]
Who states the need · A semiconductor manufacturer, not yet named. The company says the problem is heat dissipation on its chips, and a16z lists helping semiconductor manufacturers with heat dissipation among what Periodic does.[1][2] On Latent Space in October 2026 Fedus said the company has "a huge focus right now in the semiconductor industry".[9] On the job board, the Structural & Thermal computational scientist works on wafer stress and warpage, thin-film residual stress, thermo-mechanical reliability, packaging-induced stress, and fracture and delamination in semiconductor systems.[4] The editor also saw a Customer Engineer posting, to be embedded at large industrial customers in advanced semiconductor packaging, aerospace, energy and advanced manufacturing; it is no longer on today's board.[12]
Who sets the direction · Periodic and the customer's engineers together. The Forward Deployed Engineer for physics and simulation spends "significant time on-site, embedding directly with customer teams", turns customer process knowledge into computational models and drives recipe optimisation "with direct feedback loops to production"; the role requires extended time on-site in Taiwan, values Mandarin, and pays $250,000-$350,000 plus equity.[5] The semiconductor research engineer also translates "customer needs into experimental plans and results", with experience at an IDM, foundry or memory maker strongly preferred.[6]
Who runs the experiments · Two places. Periodic's on-site engineers run and calibrate the process simulations, fitting parameters to the customer's process data; the Structural & Thermal role validates models against wafer metrology, curvature and bow measurements and failure analysis.[4][5] New films and devices are made in Periodic's own Menlo Park lab, with thin-film work done at unnamed outside partner facilities until its own tools come online.[6][7]
Who validates · Not yet public. The postings describe acceptance at the customer's line: recipe optimisation feeds back "to production", and the semiconductor role builds short-loop test vehicles against "clear electrical targets".[5][6] No customer has said anything about results.
Who decides to keep paying · The customers (unnamed) and the investors. Fedus told The Generalist the company is talking about outcome-based pricing in which it keeps the IP in what its systems produce; on Latent Space he said its forward-deployed engineers train on partners' data at the customer's site "so that they can own their own intelligence".[8][9] Taken together, whether a customer buys engineering services, a model of its own or materials results is not public. On the investor side: the $300M founding round, and Forbes reporting talks on at least $500M at a $7.5B valuation, not yet announced.[2][10]
What the buyer looks at, concretely
- Heat and reliability in advanced packaging: whether wafers and packages warp, how much residual stress the film layers carry, whether thermal cycling cracks or delaminates them. The Structural & Thermal role lists exactly these, validated with wafer metrology, curvature and bow, nanoindentation and failure-analysis data.[4] For the materials themselves, this kind of buyer's acceptance language is in case 007: thermal interface materials measured for thermal resistance under ASTM D5470.[11]
- Process recipes: the on-site engineer optimises recipes, simulating processes such as microfluidics, charge transport and structural deformation. The buyer looks at what changes on the line after a recipe changes; the posting names no metric.[5]
- Devices: the semiconductor research engineer has to show new thin-film stacks meeting "clear electrical targets" in test vehicles.[6]
Not yet public · The customer; the contract's form (hours, subscription or outcome) and value; whether any customer has put a Periodic recipe or material into production; who owns IP produced in customer projects; how many customers the Taiwan work serves; who the space and defense customers are.
Decision point · The next decision is the semiconductor customer's: whether the recipe or material advice from the engineers on its site goes into its line, and how it is willing to pay. Until a customer says something about results, this chain has no acceptance on the record.
The chain's timeline
- 2025-09Launch with a $300M founding round; a16z says it works with space, defense and semiconductor customers[2]
- 2026-05Forbes: talks on at least $500M at $7.5B[10]
- 2026-06Hiring a physics and simulation forward-deployed engineer, stationed at customer sites in Taiwan[5]
- 2026-07Hiring a Structural & Thermal computational scientist for wafer warpage and packaging stress[4]
- 2026-08Hiring a semiconductor research engineer who works on customer engagements[6]
- 2026-09Fedus tells The Generalist the company is talking about outcome-based pricing, keeping the IP in the final output[8]
- 2026-10Fedus says on Latent Space that the company has a huge focus on the semiconductor industry, with forward-deployed engineers on the customer's site training on the customer's data[9]
- NextA first named customer, or a customer's own account of results. Awaiting public data.
- LaterHow it charges: outcome-based with Periodic keeping the IP, or a model sold for the customer's own use.
Sources
- Periodic Labs homepage, read 2026-11-10. periodic.com ↗ (read in full)
- a16z, "Investing in Periodic Labs", 2025-09-30. a16z.com ↗ (read in full)
- Periodic Labs job board, 32 openings, read 2026-11-10. jobs.ashbyhq.com ↗ (every posting read through Ashby's public interface)
- Posting: Computational Scientist, Structural & Thermal, published 2026-07-16. jobs.ashbyhq.com ↗ (read in full)
- Posting: Forward Deployed Engineer, Physics & Simulation, published 2026-06-03. jobs.ashbyhq.com ↗ (read in full)
- Posting: Research Engineer, Semiconductor, published 2026-08-08. jobs.ashbyhq.com ↗ (read in full)
- Posting: Postdoctoral Associate, Thin Films, published 2026-03-10. jobs.ashbyhq.com ↗ (the relevant passages read)
- The Generalist, interview transcript, 2026-09-29. generalist.com ↗ (the relevant passages read by keyword)
- Latent Space, interview transcript, 2026-10-08. latent.space ↗ (the relevant passages read by keyword)
- Forbes, 2026-05-13 (Forbes Australia's copy). forbes.com.au ↗ (read in full)
- Case 007 in this series: Entegris × Element Six × an unnamed chipmaker.
- Supplied by the editor: the description of a Customer Engineer posting; it is on neither the job board nor Ashby's public interface as of 2026-11-10, and the original was not read.