Microsoft revealed that it is developing a data center networking technology using micro LED technology with the potential to reduce energy consumption and improve performance as demand for AI and cloud computing surges.

The company forecasts that the system, created as a collaboration between Microsoft Research in Cambridge, Azure, and Microsoft 365 teams, alongside Taiwanese fabless semiconductor company MediaTek, will be commercially available with partners by late 2027.

The technology works by replacing or complementing existing fiber-optic and copper cabling used to connect servers in data centers, an area increasingly constrained by physical hurdles such as power use, distance, and reliability.

“The early concept of using LEDs to send data more cheaply and [with] lower power than both copper and fiber optics seemed like a fantasy,” Doug Burger, technical fellow and corporate VP at Microsoft Research, explained. “This breakthrough has the potential to change nearly every aspect of computing infrastructure, starting with high-bandwidth optical cables.”

Microsoft explained that, unlike traditional optical systems, which rely on lasers, the new approach uses commercially available micro LEDs combined with “imaging fiber,” a cable containing thousands of microscopic cores capable of transmitting parallel streams of light. This is said to allow data transmission across many channels simultaneously, rather than through a few high-speed channels, maintaining overall throughput while improving efficiency.

Early lab testing has indicated the micro LED system could consume about 50 percent less energy than conventional laser-based optical links. Researchers also expect it to be cheaper to manufacture and be more durable, as it is less sensitive to environmental factors like heat and dust.

Microsoft said the technology has now completed a proof-of-concept project with industry partner MediaTek and has been successfully miniaturized into a transceiver device compatible with existing data center hardware.

Vince Hu, corporate VP at MediaTek, explained that the collaboration “leverages both companies’ deep technology expertise and industry leadership and understanding of data center design to solve critical industry limitations and bottlenecks.”

Hollow core fiber designed for faster data transmission

Alongside the micro LED development, Microsoft is advancing another networking technology that transmits light through air-filled cores – known as hollow-core fiber (HCF) – instead of solid glass. This is said to enable faster data transmission and lower latency compared with conventional fiber.

“Hollow core allows us to extend that area served by one data center and an Azure region,” Frank Rey, GM of Azure hyperscale networking at Microsoft, noted. “And outside of an Azure region, if you can go a much greater distance before you need to do any signal amplification, that means fewer buildings, less power, fewer generators, less energy.”

HCF provides 47 percent faster data transmission and approximately 33 percent lower latency compared with conventional single-mode fiber, according to research from the University of Southampton Optoelectronics Research Centre, where HCF was developed. The center was later spun off as Lumenisity, which Microsoft acquired in 2022.

Microsoft currently uses HCF in some Azure regions and is in the process of being deployed in more regions globally, with manufacturing collaborations being agreed.

Microsoft last September offered a glimpse inside its $3.3 billion data center project in Mount Pleasant, Wisconsin, providing insight into the network powering what it claims to be the “world’s most powerful data center.”

The Wisconsin facility, dubbed Fairwater, has three buildings spanning some 315 acres and will house approximately 500 full-time employees. Microsoft plans to spend a further $4 billion on a similar facility in the state, with other identical data centers under construction across the US.

Unlike its existing Azure sites, which are optimized to run smaller, independent workloads, the Fairwater facility is designed to work as a singular AI supercomputer underpinned by a single, flat networking infrastructure that connects multiple pods of racks housing hundreds of thousands of Nvidia GPUs.