How is AI revolutionizing materials science research?

When the founder of Radical AI reached out to AlleyCorp’s Kevin Ryan, he made a pointed argument: material science was being overlooked in the race toward artificial intelligence. Alongside co-founder Jorge, he had grown disillusioned with the dominant trend of building interface layers on top of large models. They were looking for something more foundational. “We kept seeing wrappers on top of foundation models and thought, is this really it?” he recalled. “That can’t be the pinnacle of innovation.”

The duo spent months reading AI research, convinced the technology had potential far beyond its current uses. What they found was an opportunity at the intersection of AI and the physical world, specifically in materials science. That focus became the foundation of Radical AI, which this month announced a $55 million seed round led by RTX Ventures, with participation from NVentures (NVIDIA’s venture arm), AlleyCorp, Eni, Infinite Capital, and others.

The company is focused on developing an AI-native, vertically integrated platform for materials discovery bringing together computation, experimentation, and manufacturing into a single system. Instead of treating a material’s digital blueprint, physical synthesis, and real-world performance as separate workflows, Radical aims to merge them into what it describes as a continuous stream of scientific intelligence.

“Materials are quite large,” Joseph Krause said during a podcast conversation. “The most important spaces in the world like automotive, aerospace, defense, climate, energy, semiconductors are all a direct result of material.”

The company’s approach is structured to overcome the inefficiencies that have long plagued traditional R&D. At the center of Radical’s strategy is an autonomous lab that can run hundreds of thousands of experiments per year, generating real-world data to train and refine its AI models. This closed-loop system is designed to operate continuously, using machine learning to generate material hypotheses, test them through robotic experimentation, and update its models based on outcomes.

The concept draws inspiration from the A-Lab at Lawrence Berkeley National Laboratory, an autonomous research facility built to conduct materials experiments with minimal human intervention. One of Radical’s early collaborators helped set up that lab, and the founders flew to Berkeley to understand how such a system could be scaled beyond the academic setting. “We believed we had to bring together software, automation, and a clear opinion of how to execute,” the founder said.

Radical was initially incubated inside AlleyCorp, where it raised a $10 million pre-seed round in March 2024. The company's goal from the outset was not to improve the pace of traditional science, but to reframe it. The existing research model—hypothesis, experiment, publication, was too linear for the scale of today’s scientific challenges, the team believed. Their system would allow discoveries to unfold in parallel, drawing insights simultaneously from simulations, literature, and lab data.

A key differentiator is Radical’s control over both software and hardware. Its platform spans theoretical modeling, AI-based prediction, physical synthesis, and performance validation. That integration, the company argues, is necessary to move scientific progress out of its fragmented silos.

The $55 million seed round, one of the largest in New York startup history, is not being framed as a stealth Series A. The team sees itself in the early phase of a long execution cycle. “We still feel really early,” the founder said. “We’re super lean, hyper-focused on culture. We always ask: why does this exist the way it does?”

Over the next 18 months, Radical plans to scale its team across AI, robotics, and materials science. It also aims to complete what it describes as the most advanced materials discovery facility in the world, designed from the ground up for high-throughput experimentation and active learning.

“The data we’re collecting is the missing dataset in AI today,” the founder noted. “If we can capture and analyze it, we can feed that back into our models and truly accelerate materials innovation.”

While the technology stack is central to Radical’s mission, the company also places unusual emphasis on mindset. Internal messaging highlights the difficulty of the work, calling the pursuit “relentless” and “brutal,” and discouraging those looking for a conventional job. “We are not looking for those who simply want a job; we are looking for those who want a mission,” the company states.

Radical is not following the momentum of consumer AI tools or investing in lightweight applications. Its strategy is rooted in a belief that progress in fields like energy, aerospace, and semiconductors depends on accelerating breakthroughs in material science. That focus has drawn significant interest from corporate and financial investors alike, many of whom view the physical world as the next domain where AI will have measurable impact.

For now, the team remains focused on execution. Building infrastructure, generating data, and refining its models. “We think the most important thing at this stage is to stay focused and keep building,” the founder said.