CScale Raises $145M to Build Failure-Proof AI Optics

CScale, a Palo Alto startup building optical interconnects for large-scale AI infrastructure, exited stealth Wednesday with a $145 million Series C round backed by Nvidia and Intel Capital as first-time strategic investors, according to DCD's reporting on the round. Atreides Management, Valor Equity Partners, and Premji Invest co-led the financing, with existing investors Sutter Hill Ventures and Maverick Silicon also participating, bringing total funding to $188 million.

CScale's core bet centers on a genuinely narrow, physical engineering problem. As AI scale-up domains extend across dozens of racks, individual accelerators need extremely fast, low-latency connections to function together like one much larger computer. Optical links, which move data as pulses of light through fiber, can carry far more data over longer distances than the copper cables most AI servers currently use, but optical components include tiny lasers that inevitably fail. CScale's pitch is that it can contain those failures so they do not interrupt compute, rather than simply making the underlying components faster.

CScale CEO Martin Lund was direct about the scale of the problem across a full data center. "At that scale, reliable, predictable communication is fundamental to system economics," Lund said, according to Business Wire's release. "Lasers will fail. Compute shouldn't." Across hundreds of thousands of accelerators, failures that are rare on any single link become a continuous, fleet-level condition, degrading portions of an AI factory unless they can be contained, according to the company's own announcement.

It is worth being direct about what has and has not been disclosed. FourWeekMBA's analysis flagged a structural oddity: CScale is calling this a Series C while simultaneously exiting stealth, unusual since companies typically leave stealth earlier in their funding sequence. The company's own description offers no bandwidth figure, no latency figure, no port count, no power number, no process node, no named customer, and no shipping date beyond the general 2028 target, according to FourWeekMBA's assessment. That gap between funding stage and disclosed technical specificity is worth noting when evaluating the company's actual maturity.

Atreides Management's Gavin Baker offered a framing that positions CScale's category as increasingly central to AI infrastructure economics: "AI infrastructure is no longer just a compute problem. It is a systems problem. Every optical failure is a compute failure. At this scale, failures are not hypothetical, they are inevitable." That framing echoes the reasoning we covered in our earlier reporting on Delos Data's $100 million raise, which similarly argued that idle, network-bottlenecked compute represents genuine wasted infrastructure capital, not just an engineering inconvenience.

CScale is not alone in chasing this specific bottleneck. Rival Eliyan separately raised its own $145 million Series C to expand into electro-optical interconnects for next-generation AI systems, according to Tech Funding News's reporting on the competitive landscape. The two companies raised the identical dollar amount within a similar window, a coincidence worth noting as a sign of how much capital investors are willing to deploy into essentially the same infrastructure layer simultaneously.

For businesses evaluating AI infrastructure vendors or planning large-scale AI deployments, CScale's specific focus, keeping compute running through inevitable hardware failures rather than simply making connections faster, addresses a genuine reliability gap that directly affects uptime and total cost of ownership at scale, worth understanding as gigawatt-class AI data centers become more common.

For businesses in networking, semiconductor, or data center hardware, the presence of both Nvidia and Intel as first-time strategic investors in the same company is a notable signal, since the two chipmakers are direct competitors in other parts of the AI stack, suggesting both see genuine, shared value in solving optical reliability regardless of whose accelerators sit on either end of the connection.

What does CScale's optical interconnect technology actually do?
CScale builds an "integrated light engine" designed to provide reliable optical connectivity between AI accelerators at gigawatt scale, specifically architected to contain failures from individual lasers so they do not interrupt overall compute operations.

How much funding has CScale raised?
CScale raised $145 million in Series C funding, bringing its total funding to $188 million, with Nvidia and Intel Capital joining as the company's first strategic investors.

When does CScale expect to ship its product?
CScale has not disclosed a specific product launch date, but reporting indicates the company is targeting 2028 for shipping its optical interconnect chips.

CScale exited stealth with a $145 million Series C round backed by Nvidia and Intel Capital as first-time strategic investors, bringing total funding to $188 million, to build optical interconnects designed to keep AI compute running when individual laser components inevitably fail at scale. The Palo Alto startup, founded in 2023, has disclosed few technical specifics, including no named customer or confirmed shipping date beyond a general 2028 target, despite raising at the Series C stage while still emerging from stealth. CScale enters a genuinely crowded optical interconnect market, with rival Eliyan raising an identical $145 million Series C for a similar electro-optical approach, reflecting broad investor conviction that data movement between AI chips, not raw chip performance alone, has become a critical infrastructure bottleneck.