Orbital Industries raises $50M Series B to scale AI-designed cooling fluid and data centers

Orbital Industries raises $50M Series B for AI-designed data-centre hardware

The London-and-San-Francisco startup, formerly Orbital Materials, has closed $50M led by Plural for PFAS-free cooling fluid and modular high-density compute infrastructure. Orbital Industries, the London-and-San-Francisco AI-materials startup formerly known as Orbital Materials, has raised $50m in a Series B led by Plural with participation from Nvidia's NVentures, Radical Ventures, Compound and Fly Ventures. The funding will scale commercial deployment of the company's data-centre cooling fluid and modular compute-infrastructure products, and expand its team across London and San Francisco. The company has rebranded along with the round. Orbital Materials, founded in 2022 by chief executive Jonathan Godwin (formerly of DeepMind), chief technology officer James Gin-Pollock and chief operating officer Daniel Miodovnik, originally positioned itself around AI-discovered carbon-capture and sustainable-aviation-fuel chemistry. The company has since pivoted, or rather expanded, toward data-centre infrastructure where the immediate commercial pull has been strongest. The new Orbital Industries name captures that broader scope; the long-term ambition is to apply the same model to semiconductors, critical minerals, aerospace and energy. The two near-term products are worth describing concretely. The first is a PFAS-free dielectric cooling fluid designed for next-generation high-density GPUs. PFAS, the so-called "forever chemicals" that have anchored two-phase immersion-cooling systems in data centres for years, are facing tightening EPA and EU regulatory bans. 3M exited the production of its Novec PFAS-based coolants entirely in 2024, leaving data-centre operators scrambling for alternatives just as Nvidia's Blackwell and Rubin generations push power densities into ranges that water-cooling alone cannot handle. Orbital's fluid, developed with its AI materials platform, sits in that gap. The second product is a modular, off-site-manufactured data-centre system delivered as ready-to-deploy units. Orbital claims the approach cuts deployment timelines to as little as six months versus the traditional three years for purpose-built high-density facilities. The capacity bottleneck is real: Anthropic, OpenAI and Meta have all visibly struggled to secure enough power-and-cooling capacity to absorb their planned compute scaling, and modular pre-fabrication has emerged as the standard industry response. Orbital's differentiator, on its framing, is the AI-led engineering loop that designs the modules rather than just assembling them off-site. The underlying technical asset is Orb, the company's open-source AI model for simulating the quantum-mechanical behaviour of atoms. Published on GitHub under Apache 2.0, Orb's most recent v3 release handles fully solvated 20,000-atom enzyme simulations at scale, with the company claiming stable simulations of up to 100,000 atoms on a single GPU. Published benchmarks show Orb running three-to-six times faster than existing universal interatomic potentials and producing a 31% reduction in error against the Matbench Discovery benchmark. The AWS partnership the company has been promoting since December 2024 was struck under the old Orbital Materials name. The strategic agreement covers data-centre decarbonisation, cooling and water-utilisation technologies, with Orb itself available to AWS customers via SageMaker JumpStart and the AWS Marketplace. That commercial relationship is the primary near-term revenue anchor for the cooling-fluid product line. The Plural-led round is a meaningful endorsement on the European deeptech side. Plural partner Ian Hogarth, who led the investment, is one of the more visible UK AI-policy figures and chaired the UK AI Safety Institute's predecessor body. The Nvidia NVentures participation is the other one worth flagging: the chip giant's strategic-investment arm rarely backs hardware-infrastructure startups directly, and its involvement here signals that Orbital's cooling fluid is plausibly relevant to Blackwell-and-beyond GPU deployment. Godwin, on the company's framing, joined DeepMind around the time of AlphaFold and worked on AI-for-science and materials before leaving to found Orbital. The company has not disclosed revenue figures, customer-count specifics, or the post-money valuation implied by the Series B.

Exclusive: Orbital Industries, which uses AI to discover new materials, raises $50 million Series B | Fortune

The company, which has offices in London and San Francisco, recently rebranded from Orbital Materials. It said the new capital will scale commercial deployment of its first two products -- both of which are aimed at the data center industry -- and grow its 50-person team. It also plans to develop its AI platform for industrial applications beyond data centers. In the past two years, a number of startups have launched hoping to use AI models to discover new materials, including everything from longer-lasting batteries to biodegradable plastics to nonstick coatings that don't rely on toxic 'forever chemicals.' Among those competing with Orbital Industries are companies such as CuspAI, which raised $100 million in a Series A round in September, and Periodic Labs, which raised a $300 million seed round around the same time. But Orbital Industries is pursuing a different business model than these other startups, Jonathan Godwin, the former Google DeepMind researcher who is Orbital's cofounder and CEO, told Fortune. Rather than simply use AI models to find new chemical compounds and then license the intellectual property out to large chemicals or coatings companies such as Germany's BASF or Pittsburgh-based PPG, Orbital intends to sell the materials its AI models discover directly. It's also using AI models to optimize the manufacturing of these compounds, including building new hardware, Godwin said. In this, the company is similar to Prometheus, a startup backed by Amazon founder Jeff Bezos among others, that has raised $6.2 billion and plans to use AI to optimize manufacturing and engineering processes across a number of high-tech industries, such as aeronautics, defense, and pharmaceuticals. "We push the frontier of AI for science and engineering, but we're what's known as vertically integrated," Godwin said in an interview with Fortune. "We don't sell that software. We have hardware, manufacturing and advanced materials teams, labs and things like that, and we use that software internally to develop new advanced materials and hardware devices, and sell those hardware devices." Orbital's core AI model is called Orb, which can predict and simulates the quantum mechanical behavior of atoms. The company says Orb is the only model that can simulate 100,000 atoms on a single GPU and that it runs roughly 10 times faster than alternatives, outperforming models released by Meta and Microsoft. Orbital has made a version trained on publicly available data freely available -- including, through a multi-year partnership with Amazon Web Services announced in December 2024 -- while keeping a separate version trained on proprietary data for its own product work. Orbital used Orb to screen hundreds of thousands of candidates to find a new kind of liquid coolant that could be used to mitigate the extreme heat that server racks full of GPUs produce, but which would not use PFAS "forever chemicals" that are increasingly subject to environmental restrictions in the U.S. and Europe. The challenge, Godwin said, is that running a modern GPU rack is like packing "basically a supermarket's worth of energy into a filing cabinet." The company says it has now synthesized an entire family of molecules and is taking them through qualification with major chip providers. Orbital has also located a contract manufacturer to scale production. Godwin said developing a new cooling fluid would traditionally take "10 years and $100 million"; Orbital has done it in months for far less, he said. The cooling fluid, paired with a refrigeration system Orbital is also building, is designed to ship alongside the next generation of GPUs in 2027, Godwin said. If this timeline holds, he said, it would be the first time an AI-designed molecule has hit the commercial market in any industry. He notes that even in drug discovery, where startups have been using AI for years to search for new small molecules that can be used as pharmaceuticals, no AI-discovered drug candidate has yet to make it through clinical trials and on to the market. The path from lab to market for other kinds of new materials is often faster though since they don't have to undergo lengthy clinical trials. The company's second product is a modular data center system, manufactured off-site and delivered as ready-to-deploy units, which Orbital says can bring high-density compute capacity online in as little as six months versus up to three years for conventional builds. Both products are sold through a commercial brand, Orbital IT. Orbital was cofounded in 2022 by Godwin -- who spent five years at Google DeepMind working on AI for science and advanced materials -- alongside chief technology officer James Gin-Pollock, a repeat AI founder who previously sold a company to Shutterstock, and chief operating officer Daniel Miodovnik. Godwin has said he left DeepMind because the Google-owned lab was not going to set up an advanced-materials and hardware company itself. Ian Hogarth, a partner at Plural, said in a statement that AI progress is now constrained by "energy, heat and infrastructure" and that Orbital was "tackling those constraints directly." Godwin said part of the appeal of Plural was the firm's involvement in Proxima Fusion, the German fusion energy startup that has raised roughly $200 million in public and private capital. "There aren't so many funds that have that level of ambition for the sort of thing we could do here," he said. Plural, he added, "wants to see that happening in Europe, and I'm a Brit, I feel passionate about that too." Godwin said he ultimately wants to build the "largest industrial conglomerate" in Europe -- an AI-native counterpart to the chemicals giants that emerged a century ago. "The reason that all those companies are old is because they have really entrenched moats," he said. "In order to break down an entrenched moat, you need a pretty radical technology innovation. And that's what AI is."