San Francisco – CScale, a Silicon Valley venture focused on next‑generation AI chip interconnects, announced Wednesday that it has closed a $145 million financing round. The round was led by Atreides Management, Valor Equity Partners and Premji Invest, with participation from existing backers Sutter Hill Ventures and Maverick Silicon.
CEO Martin Lund, who departed Cisco Systems earlier this year after overseeing chip‑and‑hardware networking systems, said the new capital will speed the company’s effort to replace traditional copper cables inside AI data‑center servers with proven fiber‑optic technology. “We’re not using any exotic technologies,” Lund told Reuters. “We’re using proven technologies, integrating them together in a novel way, but preserving the same trusted fiber technology that has been out there and shipping in the industry for many, many years now.”
Why fiber matters for AI workloads
Current AI servers rely on copper interconnects to link processors. While copper is fast and reliable, its limited reach forces chip designers such as Nvidia and Advanced Micro Devices to pack processors tightly together. This dense arrangement generates significant heat, requiring elaborate liquid‑cooling systems that add cost and complexity.
Industry leaders and startups alike are pouring billions into optical solutions that use laser light instead of electrons to move data. Lasers can transmit data over longer distances with lower latency, but they are prone to failure. Most firms keep the laser separate from the chip so it can be swapped out when needed.
CScale’s approach, Lund explained, integrates the lasers directly onto the chips while ensuring that a laser failure does not interrupt data flow. The company claims this design will allow AI servers to scale more efficiently, reduce cooling requirements, and improve overall system reliability.
Funding and future milestones
The $145 million round brings CScale’s total raised capital to $188 million. With the new funding, the company plans to move from prototype to production, targeting a first shipment of its fiber‑optic AI chips by 2028. Lund emphasized that the company’s roadmap aligns with the broader industry push to modernize data‑center infrastructure as AI workloads continue to grow.
Investors see the move as a strategic bet on the long‑term demand for high‑performance AI hardware. Nvidia and Intel’s participation signals confidence that optical interconnects will become a standard component of future AI systems.
Industry context
Public companies and private startups are competing to solve the same challenge. The shift from copper to fiber is part of a larger trend in which data‑center operators invest heavily in new cooling, power and networking solutions to keep pace with the exponential growth of AI model training and inference.
By integrating lasers onto chips without sacrificing the reliability of existing fiber standards, CScale hopes to differentiate itself from rivals that rely on separate laser modules. If successful, the technology could lower total cost of ownership for AI data centers and accelerate the deployment of more powerful AI applications across sectors ranging from cloud services to autonomous vehicles.
While the company declined to share detailed technical specifications, Lund’s comments suggest that CScale’s solution builds on decades‑old fiber technology, repurposing it for the unique demands of modern AI workloads.
As the AI hardware ecosystem evolves, CScale’s financing round underscores the appetite among leading chipmakers and venture firms for innovative networking solutions that can keep the nation’s data‑center infrastructure competitive on the global stage.
Original reporting: Appleton, WI News Feed (HLL/CB) — read the source article.
