A major shift is underway in how scientists discover new materials and chemicals: instead of spending months running experiments by hand, researchers are building "self-driving labs" where artificial intelligence and robotics handle the tedious work while human scientists focus on the big-picture questions. The National Science Foundation (NSF) is backing this transformation with a four-year, $20 million grant to establish a nationwide network of these AI-enabled automated laboratories.

What Are Self-Driving Labs and How Do They Work?

Self-driving labs represent a fundamental rethinking of experimental science. Rather than researchers manually mixing chemicals, adjusting equipment, and recording results, intelligent robotic systems perform the hands-on procedures while machine learning algorithms predict which experiments to run next. The key innovation is that human scientists remain in control of the research direction and goals; the automation simply executes the work faster and more systematically.

The NSF's Programmable Cloud Laboratories program is funding a project called SPEED, which stands for Self-Driving Platforms for Experimental co-Design in Chemistry and Materials Science. NC State University is leading the effort, with partnerships from the University of North Carolina at Chapel Hill and the Massachusetts Institute of Technology.

One of the most exciting aspects of SPEED is enabling remote control of laboratory equipment. This means researchers at universities without access to expensive, cutting-edge instruments can still run sophisticated experiments by connecting to automated labs elsewhere. The promise is democratizing access to advanced research capabilities across the country.

How Much Faster Can AI-Powered Labs Accelerate Discovery?

The potential speed gains are remarkable. According to researchers involved in the project, self-driving lab modules can accelerate the process of moving from initial discovery to optimized results by 100 times or more. This acceleration comes from two complementary strategies: the robots can perform multiple experiments in parallel, and machine learning methods predict which experiments are most promising to run next, eliminating wasteful dead ends.

"The development of hardware and software tools that enable users that don't have access to advanced instrumentation to test their research ideas could be revolutionary," said Alex Miller, professor of chemistry at UNC Chapel Hill. "The individual self-driving lab modules within SPEED can accelerate the process of moving from lead discovery to optimized outcome dramatically, by 100 times or more, by performing experiments in parallel and using machine learning methods to predict the most promising next set of experiments."

Alex Miller, Professor of Chemistry at UNC Chapel Hill

This combination of speed and accessibility is reshaping what kinds of research questions become feasible. Problems that once required years of painstaking work can now be tackled in weeks or months.

What Real-World Problems Will These Labs Tackle First?

The SPEED project is focusing on two major application areas that have immediate practical value. The first is developing more efficient synthetic pathways to produce high-value specialty and fine chemicals, which are used in everything from pharmaceuticals to advanced materials. The second is accelerating the translation of next-generation materials into manufacturing for more energy-efficient digital displays and other consumer technologies.

These aren't abstract research goals; they represent bottlenecks that currently slow down innovation. Finding better ways to synthesize chemicals can reduce waste and cost. Speeding up the path from materials discovery to manufacturing can bring new technologies to market faster.

Steps to Understand How Self-Driving Labs Are Being Built

• Hardware Infrastructure: NC State has developed robotic experimentation capabilities over the past several years that form the foundation of SPEED. These systems can automatically handle chemical reactions, measure results, and adjust parameters without human intervention.
• Software and Control Systems: UNC Chapel Hill's team is developing new access and control interfaces that allow remote researchers to direct experiments and monitor results in real time, even from institutions without the physical equipment.
• Machine Learning Integration: The labs use AI algorithms to analyze experimental results and predict which combinations of variables are most likely to succeed, dramatically reducing the number of experiments needed to find optimal solutions.
• Pilot Testing: The UNC team is piloting a broad range of different chemical reactions to ensure the system works across diverse research problems, not just a narrow set of applications.

The broader context for this investment is significant. The NSF's Directorate of Technology, Innovation and Partnership is treating programmable cloud laboratories as a strategic priority, recognizing that AI-enabled automation could fundamentally reshape how the nation conducts scientific research.

This development also comes at a moment of leadership transition in the broader materials science and AI research ecosystem. Paul K. Kearns, who has served as director of Argonne National Laboratory for nearly a decade, recently announced his retirement effective March 31, 2027. During his tenure, Argonne became a driving force in harnessing AI for scientific discovery and established itself as a leader in materials science, battery innovation, and advanced computing. The lab's Aurora supercomputer, deployed in 2025 and standing among the world's first exascale machines, represents exactly the kind of computational power that enables AI-driven discovery at scale.

The convergence of these trends, self-driving labs and advanced computing infrastructure, suggests that the next phase of materials science will look fundamentally different from the past. Instead of individual researchers working in isolation, the field is moving toward networked, automated systems where AI handles routine experimentation and humans focus on creative problem-solving and strategic direction.

For researchers at smaller institutions or those without access to specialized equipment, the implications are profound. The barriers to conducting cutting-edge materials research are about to drop dramatically. A graduate student at a regional university could soon run experiments on equipment they've never physically touched, guided by AI systems that learn from thousands of previous trials.