NVIDIA just dropped AI tools that are revolutionizing how scientists discover new materials - from better OLED screens to data center cooling fluids. The company's ALCHEMI microservices, unveiled at SC25, can evaluate billions of molecular candidates up to 10,000 times faster than traditional methods, with early partners like ENEOS and Universal Display Corporation already seeing dramatic results.
NVIDIA just handed scientists a massive computational upgrade that's reshaping how we discover the materials powering tomorrow's tech. At the SC25 supercomputing conference in St. Louis, the company unveiled ALCHEMI - a suite of AI microservices that can screen billions of molecular candidates at speeds that would have seemed impossible just months ago.
The timing couldn't be better. As data centers demand better cooling solutions and display makers push for more efficient OLED screens, the race to find optimal materials has never been more urgent. Traditional computational chemistry methods simply can't keep pace with industry needs, often taking weeks or months to evaluate promising compounds.
That's where NVIDIA's new approach changes everything. Universal Display Corporation, the company behind OLED materials in everything from smartphones to VR headsets, is now evaluating molecular candidates up to 10,000 times faster than conventional CPU-based methods. "By using GPU-accelerated computing and NVIDIA ALCHEMI together with our in-house expertise, we can completely change the scale and speed of discovery," Julie Brown, UDC's executive vice president and CTO, told NVIDIA's blog.
The numbers are staggering. UDC faces a universe of roughly 10 to the 100th power possible OLED molecules - a search space so vast it would overwhelm traditional computing approaches. With ALCHEMI's AI-accelerated conformer search, they're now screening billions of candidates and reducing simulation times from days to seconds by running workloads across multiple NVIDIA GPUs in parallel.
Japanese energy giant ENEOS is seeing similar breakthroughs in their hunt for next-generation data center cooling fluids and hydrogen production catalysts. The company evaluated about 10 million liquid-immersion cooling candidates and 100 million oxygen evolution reaction candidates within just a few weeks - representing at least a 10x improvement over their previous methods.
"We hadn't considered running searches at the 10-100 million scale before, but NVIDIA ALCHEMI made it surprisingly easy to sample extensively and achieve more physically realistic results," said Takeshi Ibuka, general manager of ENEOS Holdings' AI innovation department. The speed gains aren't just about efficiency - they're fundamentally changing how scientists approach materials research by removing computational bottlenecks that previously forced narrow, conservative searches.
