While the focus of the tech world is often on AI's capabilities for generating text, images, and videos, a pioneering startup, led by a former DeepMind senior researcher, is harnessing Generative AI to innovate the production of new physical materials. Orbital Materials, founded by Jonathan Godwin, who previously contributed to DeepMind's material research initiatives, is developing an AI-driven platform designed to discover materials ranging from advanced batteries to carbon capture cells.
Godwin was motivated to launch Orbital Materials after witnessing how foundational AI techniques, like those found in AlphaFold—DeepMind’s AI that predicts a protein's 3D structure from its amino acid sequence—could revolutionize materials science. “Historically, discovering new materials involved lengthy and labor-intensive trial-and-error lab processes, often taking years before any success was evident,” Godwin explained in an email interview. “I realized that we needed a novel organization combining AI experts and materials scientists to transition materials research from simulations to practical applications.”
Creating new materials is inherently a complex endeavor, even with AI assistance. Developing specific properties—such as lightweightness combined with rigidity—requires pinpointing corresponding physical and chemical structures while determining the optimal processes (like melting or evaporating) for consistently producing these structures. Once developed, materials must undergo rigorous stress testing under various environmental conditions, such as extreme temperatures, based on their intended uses.
While AI cannot address every challenge in materials design—real-world experimentation remains irreplaceable—it can significantly conserve time and resources. By utilizing computational power, it maps out which properties and processes may yield desirable materials. “Decision-makers in chemistry and materials companies face significant challenges in product development due to the slow and costly nature of traditional advanced materials discovery methods,” Godwin noted. “However, there's been an explosive growth in demand for new advanced materials as our economies continue to electrify and reduce carbon emissions.”
Orbital Materials is not alone in applying AI to materials research and development. Companies like Osmium AI, founded by a former Google employee and backed by Y Combinator, allow industrial clients to forecast the physical properties of new materials and then optimize them through AI. Numerous academic studies have also explored accelerating material design workflows through AI combined with extensive molecular databases. DeepMind has even announced efforts to discover new materials, showcasing algorithms capable of identifying millions of potential crystal structures for future commercial use.
What differentiates Orbital Materials, according to Godwin, is its proprietary AI model tailored for materials science. “We've drawn significant inspiration from the achievements of large language models and AlphaFold to build our datasets,” Godwin shared. “The key is incorporating diverse data sources: ChatGPT, for example, utilizes everything from code to news articles and scientific literature. This variety contributes to the model's exceptional functionality.”
Named Linus, Orbital’s model serves as the core of the startup’s lab in New Jersey, where it propels material and chemical R&D. Linus has been trained on extensive datasets that include everything from batteries and semiconductors to catalysts and organic compounds. Scientists utilizing Linus can input natural language commands—like “design a material that efficiently absorbs carbon dioxide”—and the AI generates a 3D molecular structure that meets those specifications. Starting with a random cluster of atoms, Linus refines the structure through iterative processes until it produces an optimal design.
“We're adopting a full-stack AI methodology to develop a stream of materials in-house,” Godwin continued. Though Linus is not infallible—it sometimes creates materials that are physically impossible to manufacture—Godwin claims it has already successfully designed a cost-effective, reliable filter for capturing atmospheric carbon dioxide. More details about this breakthrough are expected to be announced later this year.
Based in London and supported by a team of 13 individuals, Orbital doesn't intend to manufacture the filter or any materials themselves. Instead, the company aims to advance materials to the proof-of-concept or pilot demonstration stage before collaborating with external manufacturers. Recently, Orbital raised $16 million in a Series A funding round, led by Radical Ventures, increasing its total funding to approximately $21 million. Godwin indicated that these funds will help expand the startup's data science and wet lab teams.
“Just as AlphaFold is expediting the discovery and commercialization of new drugs, the technology at Orbital Materials is accelerating the design of advanced materials at an unprecedented pace,” Godwin concluded.
Ranking
