PicoJool, which is developing AI data center interconnects based on vertical cavity surface emitting lasers, raised a $27.5M Series A led by Socratic Partners

AI cluster bandwidth costs are now subject to potential reduction, making large-scale AI deployments more accessible.

PicoJool, an AI data center interconnect startup, has secured $27.5 million in a Series A funding round led by Socratic Partners, with participation from Hudson River Trading. This funding brings PicoJool's total raised capital to $39.5 million, following a $12 million seed round in November led by Playground Global. The company is developing Vertical Cavity Surface Emitting Laser (VCSEL) technology to improve data transfer speeds and energy efficiency in AI clusters. PicoJool aims to replace traditional copper interconnects with its high-speed 100G and 200G VCSEL products and low-power MicroVCSEL configurations, offering up to 3.2 terabits per second of bandwidth. The funds will be used to expand teams in the U.S. and Taiwan, qualify products with customers, and transition from chip-level components to AOC and NPO modules, while also accelerating R&D and growing operations and sales.

The substantial Series A funding for PicoJool underscores the critical need for advanced networking solutions to support the burgeoning demands of AI workloads. As AI models grow in complexity, the bottleneck in data transfer speeds and the associated energy consumption of traditional interconnects present a significant challenge for data center operators. PicoJool's VCSEL technology offers a compelling alternative by leveraging light signals for faster, more efficient data transmission.

This investment signals a growing market appetite for specialized hardware that can alleviate the performance constraints in AI data centers. Companies operating large AI clusters, including hyperscalers and enterprises developing proprietary AI models, are likely to closely watch PicoJool's progress in commercializing its high-bandwidth, low-power interconnects. The success of this technology could redefine the cost-performance calculus for AI infrastructure and influence future data center designs.