Monday, October 26, 2026 · Edition updated 6:00 AM ETMON, OCT 26 editionNext edition in Next in --:--:--
Intelligence Verdict
AI startups · Science, materials and math

Orbital Industries

“a startup using AI to design advanced materials” Fortune

  • Nonevaluation reported
  • 2022founded
  • Has a model or API2 things shipped, by our search
The facts

Orbital Industries at a glance.

Read Oct 20, 2026. Valuations and amounts are copied as each source states them; a figure that is only reported (talks, or people familiar with the deal) is labeled as reported.

Kind of company
Science, materials and mathAI for physics, chemistry, new materials and mathematics, often with robotic labs.
Based in
United Kingdom
Founded
2022Fortune, May 28, 2026
Last closed-round valuation
No valuation reported
What it has shipped
Has a model or API
Founders
  • Jonathan Godwin: cofounder and CEO; former Google DeepMind researcher, five years on AI for science and advanced materials (Fortune, May 28, 2026)
  • James Gin-Pollock: cofounder and CTO; repeat AI founder who sold a company to Shutterstock (Fortune, May 28, 2026)
  • Daniel Miodovnik: cofounder and COO (Fortune, May 28, 2026)
Latest news

Raised a $50M Series B led by Plural; rebranded from Orbital Materials; says a PFAS-free data-center coolant is slated to ship with next-generation GPUs in 2027. May 28, 2026

Fortune, May 28, 2026

What it has built

Shipped, or not yet.

“Has a model or API” is what our search found: something the public can use or buy, a model or API, research only, or nothing public.

open-source · Apr 10, 2025

Orb-v3 interatomic potential models; code and weights on GitHub under Apache 2.0

Source
paper · Apr 10, 2025

Orb-v3: atomistic simulation at scale (preprint)

Source
1 round

Funding, as reported.

Every round we found, with the amount, valuation and investors as the source states them.

WhenRoundAmountValuationInvestors namedSource
May 28, 2026Series B$50 millionNot statedPlural (lead), NVentures, Radical Ventures, Compound, Fly VenturesFortune
Outside checks

What others have found.

  • UniFFBench, by researchers at IIT Delhi, Intel Labs and NVIDIA, benchmarked six force-field models including Orb against experimental data; Orb "exhibits failure across all elastic tensor components, with MAPE values consistently exceeding 80%" while doing well on structural metrics (preprint). (arXiv 2508.05762 (IIT Delhi, Intel Labs, NVIDIA authors), Aug 2025)

Our note: Fortune says the company has offices in London and San Francisco and names no single HQ; country set to UK on the London base. It has a multi-year AWS partnership announced Dec 2024 (per Fortune). Customers are unnamed.