
Silicon Valley-based startup CScale has raised $145 million in venture capital to develop optical connectivity technology for artificial intelligence data-centre infrastructure, with Nvidia and Intel participating as strategic investors.
The funding comes as AI infrastructure companies increasingly explore optical connections as an alternative to copper links used to connect chips inside data-centre servers. CScale is developing technology that uses fibre-optic connections to link AI chips and aims to address reliability challenges associated with optical networking.
The company, formerly known as CSpeed, is led by CEO Martin Lund, who earlier this year left Cisco Systems, where he had been responsible for chip and hardware systems. CScale has disclosed limited technical details about its technology as it emerges from stealth.
The latest financing round was led by Atreides Management, Valor Equity Partners and Premji Invest. Existing investors Sutter Hill Ventures and Maverick Silicon also participated in the round. Nvidia and Intel joined the financing as strategic investors.
CScale said its technology is designed to replace copper connections between AI chips with optical links. Copper connections are widely used because they are fast and reliable, but their limited reach can restrict how closely and efficiently processors can be connected inside AI servers.
These limitations have contributed to chip designers such as Nvidia and Advanced Micro Devices placing processors closer together within servers. The higher concentration of computing hardware can increase heat generation and require sophisticated liquid-cooling systems.
Optical connections transmit data using flashes of laser light rather than electrons. Companies across the semiconductor and data-centre industries are investing in optical technologies, although laser reliability remains a challenge. CScale is working on an approach that integrates lasers onto chips while allowing data transmission to continue if a laser fails.
Lund said the company is using established technologies and combining them in a different way.
“We’re not using any exotic technologies,” Lund said. “We’re using proven technologies, integrating them together in a novel way, but preserving the same trusted fibre technology that has been out there and shipping in the industry for many, many years now.”
CScale plans to ship its chips by 2028. The company has raised $188 million in total funding to date, including the latest financing. The new capital will support the development and commercialisation of its optical interconnect technology for AI infrastructure.
CScale’s technology is aimed at the growing requirements of AI computing systems, where large numbers of accelerators need high-bandwidth and low-latency connections to operate together. The company is developing optical interconnect technology intended to maintain connectivity even when individual optical components experience failures.
The funding gives CScale additional backing from major semiconductor companies as it works toward commercialising its technology. With Nvidia and Intel among its strategic investors, the startup is developing its optical networking platform for the next generation of AI data-centre infrastructure.




