Microsoft Build 2025: How Microsoft Used Its New Platform ‘Discovery’ and Innovated a Coolant

Just before closing his marathon Build 2025 keynote recently, Microsoft CEO Satya Nadella talked about the ‘next frontier in science.’ He officially launched Microsoft Discovery, its new enterprise agentic platform to accelerate research and development (R&D).

Giving a parallel, Nadella said, “Microsoft's AI platform empowers developers and knowledge workers, and one of the most exciting things in the years to come will be real breakthroughs in the scientific process itself. It's about accelerating how we create new materials, new compounds, and new molecules.”

Making that into a reality, Microsoft is applying the entire AI stack, right from infrastructure to agents to scientific workflows. This vision culminates in the launch of its Discovery platform. The company is positioning it alongside GitHub (for developers) and Microsoft 365 with Copilot (for knowledge work). Microsoft Discovery is for science. A great ecosystem indeed.

What is Microsoft Discovery and Why It’s Disruptive?

It’s built on a graph-based knowledge engine. What it means is, it uses Graph RAG (Retrieval-Augmented Generation). Hence, it goes beyond fact retrieval, as it understands nuanced scientific knowledge across public scientific data and private enterprise data (e.g., for biopharma research).

It is also integrated with Foundry, which is Microsoft's platform for building and orchestrating agents. It empowers specialized R&D agents to:

• Collaborate iteratively
• Generate, simulate, and refine scientific candidates
• Learn from each research cycle

Nadella said, “Think of Copilot for coding or knowledge work—this is the science agent.”

According to Microsoft, Discovery is a foundational platform for transforming the scientific method through agentic AI, unlocking faster, more integrated, and intelligent scientific discovery.

Microsoft Discovery: A Live Use Case on Stage

To underscore the power of this platform, the Microsoft CEO called John Link, Principal Program Manager, Product Innovation and also the Chemistry Lead, to showcase how radical the company’s Discovery platform is for advancing innovation and new discoveries in science.

AI Agents Making Scientific Discovery Real

In a demo, John showed how he led a team of AI agents to discover a new, environmentally friendly immersion coolant—an emerging area in data center cooling.

Not just proposed, but synthesized. Not by accident, but by design. And the scientist behind it? A team of AI agents, leveraging Microsoft’s Science Platform.

John said, “Most immersion coolants used in data centers today are based on PFAS, or ‘forever chemicals,’” John explained. “They're effective, but environmentally harmful. The question was: Could we find a better alternative?”

By the way, if you are wondering what PFAS is, it stands for ‘Per- and Polyfluoroalkyl Substances.’ It is a large group of man-made chemicals that have been used in industry and consumer products since the 1940s. They are called "forever chemicals" because they don’t break down in the environment or human body. And they’re found in nonstick products, water-repellent clothing, stain-resistant fabrics, firefighting foam, and some electronics coolants.

Making of a Coolant with Microsoft Discovery

Let’s look at the key steps John went through:

Step 1: Reason Over Knowledge - To begin with, John asked the platform to research coolants and their properties to identify starting points. What came back wasn’t just a search result—it was a deep dive.

A network of AI agents reasoned over both public and internal scientific research. Behind the scenes, a knowledge graph connected concepts, experiments, and data points in ways large language models can’t.

“It produced a summary on the left, and a detailed, citation-linked report on the right. I could validate and iterate before moving forward,” John said.

Step 2: Hypothesis Generation - Armed with insight, John asked the agents to generate a plan specific to the problem—without writing a line of code. “All I told it was what I needed, like a target boiling point and dielectric constant—and it took over from there,” he said.

The result? A fully automated workflow:

• Generative chemistry to create millions of novel molecules
• AI models to screen for promising candidates
• HPC simulations for validation

“I could modify the steps if I wanted. But the plan looked solid, so we ran with it.”

Step 3: Experimentation - This is where things got real. Microsoft Discovery executed the plan, allocating the best HPC resources in Azure to simulate the behavior of these new compounds. In the future, this same pipeline will incorporate quantum computing to further speed discovery.

“What used to take years of lab work now happens in hours,” John said. “It’s like having an AI research lab on demand.”

Final Outcome: A Discovery

At the end of the loop, one candidate stood out. The team synthesized it in the lab.

John played a video: a PC running Forza Motorsport submerged in the new coolant. No fans. No overheating. “It’s literally very cool,” he said.

This wasn’t just a demo. It was a proof of concept, for how science itself is being redefined.

The Future of Scientific Discovery: A Platform for Every Scientist

“We discovered a PFAS-free coolant in days. But imagine using this to design new therapeutics, semiconductors, or advanced materials,” John concluded.

The implications are enormous and far-reaching. Imagine scenarios like faster climate solutions, accelerated drug discovery, and scientific access for teams that once lacked the resources to compete. In the end, Microsoft didn’t just showcase an AI platform. It showed a glimpse of what the future of science could be with technology—the endless possibilities, where AI is a co-researcher, innovator, accelerator, and a gamechanger.

“We found a promising coolant without forever chemicals. But think bigger—imagine using Microsoft Discovery to design new therapeutics, semiconductors, or advanced materials," said John.

To sum aptly: “It worked for us. The next great breakthrough is yours to discover,” said John.

Also Read:

Microsoft Build 2025: The Rise of the Agentic Web and Other Key Takeaways

Microsoft Turns 50: Tracing Its Journey from Code to Cloud Platforms

Why Microsoft is Shuttering Skype: Once an Internet Icon, Now History

Microsoft’s Majorana 1: A Quantum Leap in Computing