Radical AI says its robotic laboratory can design and test hundreds of new alloy compositions in weeks, helping US aerospace and defence manufacturers reduce exposure to supply-constrained metals
As the United States seeks to reduce its dependence on China for critical minerals, New York-based Radical AI is pursuing an alternative strategy: redesigning advanced materials so they require less of the metals most vulnerable to supply disruptions.
The company has developed an autonomous laboratory platform that combines artificial intelligence and robotics to design, manufacture and test new alloys. It says the approach can sharply reduce the time needed to discover materials for aerospace, defence and other industrial applications.
At its facility in Brooklyn Navy Yard, Radical AI has produced 300 new alloy compositions in approximately 16 weeks, chief executive Joseph Frause told MINING.COM.
The company’s work is backed by RTX and Nvidia. In September, it also partnered with Sandia National Laboratories to study metal-hydrogen interactions and identify materials that could separate hydrogen from waste-gas streams.
Radical AI says its technology could give US manufacturers another way to address supply-chain vulnerabilities: instead of relying solely on securing more supplies of scarce metals, companies could develop materials that deliver the required performance using more readily available elements.
“If you have a system that can design new materials based on the constraints that you give it, and you have existing materials that have real constraints from the supply chain, we can help solve that problem,” Frause said.
“We can design new materials that take into consideration those constraints while still executing on the function or what we call the property set that is required for an application.”
Redesigning alloys to reduce reliance on critical minerals
Washington’s critical-minerals strategy has largely focused on expanding domestic mining, processing and refining capacity to reduce reliance on foreign suppliers.
Radical AI operates further downstream. It uses refined metals to develop advanced materials tailored to industrial performance requirements, rather than mining or processing the minerals itself.
Its platform can identify combinations of elements that deliver specific properties while reducing exposure to metals affected by supply constraints.
One target is hafnium, a strategically important element used in high-temperature alloys for aerospace and defence applications.
Radical AI says it has developed high-performance alloys designed for extreme temperatures that contain very little hafnium, and in some cases almost none.
“Hafnium is one of those elements in the periodic table that is majorly owned from China’s supply chain and is a risk factor to the Department of War and just the United States and our allies,” Frause said.
Manufacturers can specify which elements they want excluded from a new material while retaining required properties such as strength, durability and resistance to heat and oxidation.
“We are mixing up the ingredients together, not messing with the ingredients,” Frause said. “We are using the ingredients that are available. Now, what we can do is decide what ingredients should be on the list.”
The approach could help manufacturers develop alternatives to alloys containing metals vulnerable to export restrictions, geopolitical tensions or limited domestic production.
AI-driven experiments aim to compress decades of materials research
Traditional materials development can be lengthy and expensive. The source report says mining projects can take 10 to 15 years to reach operation after discovery, while developing a single material system can cost more than $100 million.
The result is that some aerospace and defence manufacturers continue to rely on alloys developed decades ago.
“A lot of the advanced materials that they use, the structural metals they use, are old. 20, 30, 40, 50 years old in some of the applications,” Frause said.
Radical AI aims to shorten the process by linking AI-based design directly to physical experimentation.
Once the system identifies a promising composition, it develops an experimental plan to synthesise and evaluate the alloy. The laboratory then produces a physical sample weighing about 50 to 200 grams.
The sample is analysed for chemical composition, grain structure and mechanical properties. Tests can include hardness, tensile strength and performance at high temperatures, including conditions relevant to hypersonic applications.
The experimental results are then fed back into the machine-learning models to inform the next round of designs.
“We’ll take all of that data and we’ll analyze it with machine learning and then we will feed it back into that design loop,” Frause said.
“We call this the active learning loop so that the next time I make a new alloy, I am looking at the past results as I’m doing so.”
By continuously updating its models with laboratory data, the company hopes to accelerate a process traditionally carried out through repeated experiments by materials scientists.
“Put simply, you take a 20-year development timeline and you shrink it down to one, two, three years,” Frause said.
Materials designed alongside aircraft and defence systems
Radical AI says its technology could also change when and how manufacturers develop materials.
In conventional product development, companies often select from existing alloys. The startup’s approach would allow manufacturers to design materials specifically for the needs of a new aircraft, engine, defence system or industrial machine.
The model, known as concurrent engineering, would allow materials research to proceed alongside product design.
“As I am designing my new airplane, I am designing the materials that go in my airplane,” Frause said.
“That does not happen in the alloy space today.”
Developing application-specific materials could help manufacturers pursue performance improvements that existing alloys cannot deliver. It could also allow them to account for the availability of critical minerals at the start of a project, rather than discovering supply constraints after a design has been finalised.
Radical AI sells its discovery platform, not commercial quantities of alloys
Founded in January 2024, Radical AI says it has commercialised its autonomous laboratory platform and already has paying customers.
The company does not produce alloys at commercial scale or enter into offtake agreements. Customers use its platform to accelerate materials discovery and can retain intellectual property rights to compositions developed through the system.
“Our product is the platform, the self-driving lab. That’s what we’ve commercialized and that’s what we do have customers on today,” Frause said. “We don’t off-take the materials.”
The distinction matters because alloys discovered through the platform still need further development and qualification before they can be deployed commercially.
Aerospace and defence materials, in particular, must meet demanding safety, reliability and performance standards. A promising laboratory result does not by itself establish that a material is ready for use in aircraft, engines or defence systems.
AI-driven discovery is not expected to eliminate the need for mining, processing or refining critical minerals. But materials substitution could give manufacturers more flexibility in choosing which metals they use.
For mining companies, the technology could eventually affect demand for some critical metals, even as new industrial applications create opportunities for others.
“It’s really the way we do science that’s going to impact the entirety of the materials industry at large,” Frause said. “It’s going to be a new way to do scientific discovery.”