Microsoft is reimagining digital plumbing to meet the surging demands of AI and cloud computing. Researchers at the Microsoft Research Lab in Cambridge have unveiled a MicroLED-based cabling system designed to outperform traditional copper and laser-based fiber optics. By using inexpensive MicroLEDs—tiny light sources often found in modern screens—and imaging fiber typically used in medical endoscopes, this system can transmit data across thousands of parallel channels simultaneously.

This wide and slow architecture mimics a broad river rather than a narrow, high-pressure stream, allowing for massive data throughput with 50% less energy consumption. Unlike copper, which is physically limited to short distances of about two meters, or standard fiber, which is sensitive to temperature fluctuations and dust, MicroLEDs offer a resilient, high-bandwidth solution for connecting the graphics processing units (GPUs) that power AI models.

Complementing this is Hollow Core Fiber (HCF), which transmits light through air-filled cores instead of solid glass. Because light travels faster through air, this innovation reduces latency (the delay before data transfer begins) by 33% and boosts speed by 47%. These hardware advancements provide the necessary physical foundation to support the next generation of AI model growth and global cloud infrastructure efficiency.

Microsoft is reimagining digital plumbing to meet the surging demands of AI and cloud computing. Researchers at the Microsoft Research Lab in Cambridge have unveiled a MicroLED-based cabling system designed to outperform traditional copper and laser-based fiber optics. By using inexpensive MicroLEDs—tiny light sources often found in modern screens—and imaging fiber typically used in medical endoscopes, this system can transmit data across thousands of parallel channels simultaneously.

This wide and slow architecture mimics a broad river rather than a narrow, high-pressure stream, allowing for massive data throughput with 50% less energy consumption. Unlike copper, which is physically limited to short distances of about two meters, or standard fiber, which is sensitive to temperature fluctuations and dust, MicroLEDs offer a resilient, high-bandwidth solution for connecting the graphics processing units (GPUs) that power AI models.

Complementing this is Hollow Core Fiber (HCF), which transmits light through air-filled cores instead of solid glass. Because light travels faster through air, this innovation reduces latency (the delay before data transfer begins) by 33% and boosts speed by 47%. These hardware advancements provide the necessary physical foundation to support the next generation of AI model growth and global cloud infrastructure efficiency.