Orbital Industries has turned a materials science bet into a real business, and investors are paying attention.
The London and San Francisco startup has raised a 50 million dollar Series B led by Plural, with Nvidia's venture arm NVentures, Radical Ventures, Compound, and Fly Ventures also joining the round, according to Fortune. The timing matters. AI-for-science is moving from research curiosity to commercial race, and Orbital is trying to prove that advanced materials can be discovered, manufactured, and sold faster than the old industrial cycle allows.
The company recently rebranded from Orbital Materials, a small but telling move that signals broader ambition. It is not just trying to build software for scientists. It wants to become an AI-native industrial company, one that uses its models internally to design physical products and then sells those products directly through a commercial brand called Orbital IT. That is a different playbook from the licensing-heavy model many deep tech startups prefer, and it puts more pressure on execution, manufacturing, and distribution.
Jonathan Godwin, the former Google DeepMind researcher who co-founded the company, told Fortune that Orbital is targeting the bottlenecks around energy, heat, and infrastructure. That is where the first products are aimed. The startup says it has used its core model, Orb, to identify a new liquid coolant for GPU racks and a modular data center system that can be deployed faster than conventional builds. Those are not abstract lab goals. They sit in one of the biggest real-world pain points of the AI boom, which is how to keep dense compute clusters running without turning every data center into a thermal problem.
Orbital's approach is more vertically integrated than most AI materials startups. The company is not simply discovering compounds and licensing the intellectual property to chemicals giants such as BASF or PPG. Instead, it wants to discover the material, optimize the manufacturing process, build the hardware around it, and sell the finished system itself. That keeps more value inside the company, but it also means Orbital has to do more than look clever in a deck. It has to ship, qualify, and scale.
Orbital says its latest Orb models can bring quantum-level accuracy to simulations involving tens or even hundreds of thousands of atoms, while its sparse direct model is 10 times faster and 5 times more memory efficient than MACE-MPA-0, and 40 times faster and 40 times more memory efficient than Sevennet. It has also made model weights and code available under an Apache 2.0 license. That split between openness and product focus is important. It lets the company build credibility in the broader scientific community while reserving its strongest commercial work for its own industrial stack.
Orbital is not alone in the race. CuspAI raised 100 million dollars in September 2025, while Periodic Labs announced a 300 million dollar seed round later that month. The field is getting crowded because the market opportunity is obvious. If AI can shorten materials development from years to months, the payoff could touch semiconductors, batteries, superconductors, coatings, and industrial chemistry. The hard part is that those categories are notoriously slow to validate. A model can suggest a promising compound. Industry still has to prove it works in the real world.
That is where Orbital's mix of AI, hardware, and manufacturing starts to look more interesting than a pure software pitch. The company told Fortune that one of its cooling fluid programs, paired with a refrigeration system it is also building, is designed to ship alongside next-generation GPUs in 2027. It also says a conventional development cycle for a new cooling fluid could take 10 years and cost 100 million dollars, while Orbital has done it in months for far less. Those are bold claims, but the commercial logic is straightforward. Whoever solves heat and infrastructure for AI clusters stands to sell into one of the most urgent supply chain constraints in the industry.
The capital is chasing science
This round also fits a broader pattern in venture capital. Investors are showing more appetite for AI that touches physical systems, not just software workflows. In the same way drug discovery has attracted money because models can compress long research cycles, materials science is drawing interest because it sits at the intersection of compute, manufacturing, and industrial demand. Orbital's pitch is that it can use the same core AI advantage to unlock products that matter in data centers first, then move outward into semiconductors, energy, and other physical industries.
Godwin has said he wants to build the largest industrial conglomerate in Europe, an AI-native counterpart to the chemicals giants that emerged a century ago. That ambition may sound grand, but it is also a useful frame for what Orbital is actually doing. The company is not selling a model alone. It is trying to build a new kind of industrial stack, one where software, lab work, and manufacturing are fused into a single operating model. If that works, the startup will not just be another AI-for-science name in a crowded market. It will have shown that the real value in materials discovery is not finding molecules faster, but turning them into products people can actually buy.
Also read: Geordie AI's 30 million raise shows the agent stack is getting a control layer • SpaceX Starship grounding adds fresh risk to its valuation story • Richard Liu draws a line on robots and workers at JD.com
