Microsoft enlisted 3M to supply interconnect technologies to reduce the need for cleaning connectors across its cloud and AI infrastructure.
Traditional connectors use physical contact between fibers, allowing light to travel from one fiber to another. The challenge is that such connections require frequent cleaning and inspection as the tiniest bit of dust can disrupt interconnections.
3M’s expanded beam optical (EBO) technology does pretty much what its name implies, increasing the beam area to overcome the effects of surface particles. The reduced sensitivity to dust means engineers can insert interconnect cables without having to worry about contamination, a benefit 3M contends will help the hyperscaler more quickly install fiber connections.
The optical interconnect can be configured and scaled from 12 to 144 fibers, supporting both rack and infrastructure applications.
“3M’s EBO solution will help unlock new levels of performance, reliability, and efficiency to ensure customers can run their cloud and AI workloads on a trusted, sustainable and advanced environment,” Cliff Henson, Microsoft’s corporate VP for cloud supply chain and engineering, noted.
Like its hyperscale rivals, Microsoft has aggressively built out its cloud and AI infrastructure, including the mammoth scale-across supporting Fairwater facilities. The bigger the stack, the more networking the hyperscaler will have to install and manage, so reducing dusting times means engineering teams can focus on getting the best out of their graphic processing units (GPUs).
Microsoft’s early use of 3M’s EBO tech showed the potential to reduce network deployment timelines while also demonstrating strong signal performance in live data center conditions.
“At 3M, we view AI as a powerful tool that can accelerate growth, improve customer experiences, and help our teams work more effectively,” 3M EVP and Chief Strategy Officer (CSO) Jon Van Wyck added. “Our collaboration with Microsoft supports that vision through targeted optimization opportunities for our enterprise while advancing the infrastructure needed to power the future of AI.”