Orbital Industries, a London-and-San-Francisco startup, just closed a $50 million Series B to do something that has not been done in commercial hardware yet: ship an AI-designed molecule into production. The cooling fluid the company engineered using its Orb simulation platform is scheduled to deploy alongside next-generation GPUs in 2027. If that timeline holds, it is the first time a molecule designed by artificial intelligence will move from lab simulation to factory floor at scale. The round was led by Plural, with NVIDIA NVentures, Radical Ventures, Compound, and Fly Ventures participating as co-investors. That is not decorative: the world's largest GPU maker is betting its own capital that this particular path to materials discovery works faster than the traditional one.
Orbital's thesis rests on simulation speed. The company claims its Orb engine can model the behavior of 100,000 atoms on a single GPU, running molecular dynamics simulations three to six times faster than existing universal interatomic potentials and producing a 31 percent reduction in error against the Matbench Discovery benchmark. Microsoft and Meta both use this class of tool for materials research; Orbital's claim is that it beats both on speed and accuracy. That matters because materials discovery traditionally works like this: chemists propose a molecular structure, synthesize it, test it, watch it fail, then repeat for months or years. Simulation compresses that loop. If Orbital's engine genuinely runs ten times faster than competitors, the cycle time moves from years to weeks, which changes who can afford to iterate and, more importantly, who can afford to customize materials for specific hardware.
The initial product is a PFAS-free dielectric cooling fluid designed for high-density GPU clusters. Traditional cooling systems rely on perfluorocarbons, which carry regulatory and environmental friction. Orbital designed a replacement using its Orb engine, and AWS has already agreed to a multi-year partnership to test the system. The timing is the tell: the fluid needs to ship with the next GPU generation in 2027, which means Orbital has roughly twelve months to move from prototype to manufacturing qualification. That is not impossible, materials used in semiconductor cooling have well-established regulatory pathways, but it is a hard target. The company is not just discovering materials; it is vertically integrating manufacturing. Most AI-materials startups license IP to established suppliers. Orbital is building refrigeration systems and factory-ready modular data centres alongside the fluid itself. That means the company owns the full stack from simulation to deployment, which is where the margin hides.
Competitive advantage here is not durable, but the window is real. NVIDIA, Samsung, and Intel all run internal materials research programs. If Orbital proves that AI-accelerated simulation actually reaches production faster than traditional chemistry, the GPU makers will absorb the capability in-house. The Series B capital is built on the assumption that Orbital can stay ahead of that internalization, that the company can move from cooling fluids to other bottleneck materials (adhesives, underfills, thermal interface materials) fast enough to build a moat. The modular data centre angle suggests the company is thinking about this differently: instead of licensing molecules to OEMs, Orbital could become an infrastructure vendor, bundling custom materials, cooling systems, and factory-built pods into a single product. That path keeps Orbital downstream longer.
Watch three specific markers to see whether the bet pays off. First, the 2027 GPU deployment, does the cooling fluid actually hit production hardware, or does it get pushed to 2028 for regulatory or manufacturing reasons? Second, comparative performance: does the fluid actually cool GPUs more efficiently than incumbent fluids at the same cost, or is it a premium-only solution? Third, scalability in manufacturing: as AWS and other customers ramp usage, does Orbital's refrigeration system scale to thousands of units per year without quality degradation? The simulation engine is the science. The manufacturing and deployment are the market test. Right now, Orbital has built a narrative where AI-driven materials discovery is faster than empirical chemistry. The Series B money is betting that narrative holds under real production constraints.