AI Affairs, home

Saturday 3 October 2026

Markets

CScale raises $145 million for AI optical links with Nvidia and Intel backing

The Series C brings total funding to $188 million. CScale plans to develop optical connections designed to keep large AI computing systems running when individual links fail.

Nvidia headquarters beneath a large white lattice canopy in Santa Clara
Photo: Coolcaesar, CC BY-SA 4.0, via Wikimedia Commons (cropped)

CScale emerged from stealth on 30 September with a $145 million Series C round backed by Nvidia and Intel Capital, bringing its total funding to $188 million, Citybiz reported.

Key points

  • Atreides Management, Valor Equity Partners and Premji Invest co-led the Series C.
  • Nvidia and Intel Capital joined as CScale’s first strategic investors.
  • CScale is developing an optical light engine intended to contain link failures in large AI computing systems.

Three firms co-lead CScale’s Series C

Atreides Management, Valor Equity Partners and Premji Invest co-led the financing. Sutter Hill Ventures, which has backed CScale since its inception, and existing investor Maverick Silicon also took part. Nvidia and Intel Capital joined the round as the company’s first strategic investors.

The Palo Alto, California-based company plans to put the capital towards product development and broader commercialisation of its optical interconnect technology. Its product is aimed at AI scale-up systems, which connect accelerators so they can operate together as one computing system.

The participation of Nvidia and Intel Capital places investors associated with chipmaking alongside the three firms leading the round. CScale is building connections for systems containing GPUs and other accelerators, rather than the processors themselves.

CScale targets failures across dozens of racks

CScale is developing an integrated optical light engine for high-bandwidth, low-latency connections. It is targeting future gigawatt-scale AI data centres, where it expects scale-up systems to stretch beyond individual racks.

The company argues that a system with thousands of closely connected accelerators across dozens of racks faces a different reliability problem from a single optical link. Failures that are uncommon on one link can recur across a much larger computing fleet, it says.

CScale says its architecture is designed to contain an optical failure while the wider computing system continues to run. Its approach combines photonics, electronics and software, with reliability across the system taking priority over making an individual failed component easier to replace.

Chief executive Martin Lund distinguished between replacing a faulty part and maintaining service while it fails. “Lasers will fail. Compute shouldn’t,” he said, according to Citybiz. CScale’s design is intended to preserve that continuity as connected systems grow.

Gavin Baker, managing partner and chief investment officer at round co-leader Atreides Management, said bandwidth alone would be insufficient as scale-up systems move towards gigawatt-class deployments. He identified system reliability as a requirement alongside the capacity of the connections between accelerators.

The product’s commercial case rests on that same deployment pattern: larger groups of accelerators connected across more racks, with optical faults occurring somewhere in the fleet. CScale’s stated design goal is to keep those faults from interrupting the broader system.

Lund and Kohli bring silicon backgrounds

Lund previously led Cisco’s Common Hardware Group. His responsibilities there covered silicon, hardware systems and optics, including the Silicon One networking architecture. He also held senior positions at Microsoft, Cadence and Broadcom.

Founder and chief technology officer Sanjai Kohli previously co-founded GPS semiconductor company SiRF. He later founded Inovi, which Facebook acquired in 2014. The two executives’ backgrounds span the networking and silicon work involved in CScale’s proposed interconnect.

CScale was founded in 2023 and has approximately 85 employees globally, Citybiz reported.

Topics: Chips, Data centres, Funding