Microsoft’s light-based computer could make AI 100x more efficient

A new kind of computer that harnesses light instead of digital switches could dramatically reduce the energy demands of artificial intelligence (AI), according to a groundbreaking study published Sept. 3 in Nature. Microsoft researchers behind the project describe it as a "new computing paradigm" with the potential to reshape the future of computing.

A Prototype Optical Machine

The team has built an analog optical computer (AOC) that uses micro-LEDs and camera sensors to perform calculations. Unlike traditional digital computers, which rely on billions of transistors flipping on and off, the AOC manipulates light and varying voltages in a feedback loop to perform additions and multiplications. The machine repeatedly computes problems, improving each pass, until it reaches a stable final answer.

The result is astonishing energy efficiency.

“The most important aspect the AOC delivers is that we estimate around a hundred times improvement in energy efficiency,” said Jannes Gladrow, a Microsoft AI researcher and co-author of the study. “That alone is unheard of in hardware.”

Specialized Power for AI and Optimization

The researchers caution that the AOC is not a universal computer. Instead, it is designed as a “steady-state finder” — making it well suited for AI workloads and optimization challenges but not for general-purpose computing.

Still, within its scope, the system shows huge promise. In early experiments, the AOC:

• Classified images as accurately as digital computers during simple machine learning tasks.
• Reconstructed a 320×320-pixel brain scan using only 62.5% of the data, suggesting potential applications in faster MRI scans.
• Outperformed existing quantum computers at solving financial optimization problems involving fund exchanges and risk minimization.

Because the AOC does not convert analog signals into digital ones mid-process, it avoids the energy losses and speed bottlenecks inherent to conventional architectures.

The Role of the Digital Twin

Alongside the physical machine, the researchers developed a “digital twin” — a software model that simulates the AOC’s operations. This allows them to experiment with larger, more complex problems than the prototype can yet handle.

“The digital twin is where we can work on larger problems than the instrument itself can tackle right now,” explained Michael Hansen, senior director of biomedical signal processing at Microsoft Health Futures.

The Future of Light-Based Computing

For now, the AOC remains a prototype, limited in the scale of problems it can solve. But researchers envision next-generation systems packed with millions or even billions of micro-LEDs, enabling unprecedented computational power with dramatically lower energy consumption.

“Our goal, our long-term vision is this being a significant part of the future of computing,” said Hitesh Ballani of Microsoft’s Cloud Systems Futures team.

If realized, this light-driven technology could help address the ballooning energy demands of AI — ushering in a future where computation is not only faster, but far more sustainable.