CScale has raised $145 million to develop optical networking technology for AI infrastructure, according to a report by CXO Digitalpulse. The round lands amid a surge of investor capital flowing into the interconnect layer of AI data centers — the networking fabric that moves data between GPUs and memory, and that increasingly constrains how large clusters perform.

The $145 million figure is the confirmed core of the announcement. The report did not specify the round's lead investors, valuation, or a staged breakdown of the capital. What the financing signals is unambiguous: private markets are now underwriting startups that attack the data-movement problem inside AI clusters rather than the compute itself.

Why is optical networking drawing this level of funding?

AI training clusters have scaled to tens of thousands of accelerators, and the links between them have become a first-order cost and performance constraint. Copper interconnects face distance, power, and density limits as rack power climbs, pushing operators toward optical links for switch-to-switch and, increasingly, chip-to-chip connections.

That shift has opened commercial space for suppliers of optical components, co-packaged optics, and optical interconnect silicon. Established players including Broadcom, Marvell, and Coherent have built large businesses in this segment; Nvidia's roadmap has also moved toward optical interconnects for its highest-end platforms. A new entrant such as CScale is betting that AI-specific optical networking architectures warrant purpose-built approaches rather than adapted telecom parts.

What is CScale actually building?

The source report confirms the funding amount and the company's focus on optical networking for AI infrastructure. It does not detail product specifications, wafer or manufacturing partners, customer engagements, or shipment timelines. Any claims about specific process nodes, data rates, or design wins for CScale would go beyond the confirmed record.

The category itself is well defined, though. Optical networking for AI spans:

• High-speed transceivers linking switches across cluster fabrics
• Co-packaged or near-package optics that shorten electrical paths
• Optical circuit switches that reconfigure fabric topology
• Laser sources and waveguide integration for dense rack-scale links

Which of these CScale targets first remains undisclosed.

How crowded is the field?

The financing arrives during a broader wave of capital directed at AI infrastructure internals. Optical-module suppliers have posted rapid revenue growth as hyperscalers build out cluster capacity, and component makers across the supply chain — from lasers to connectors to silicon photonics foundries — have attracted both investment and multi-year supply agreements.

That demand is driven by concrete buildouts. Every large GPU cluster requires thousands of optical links, and per-link data rates have climbed from 400G to 800G with 1.6T platforms ramping next. Startups entering now must compete not only with incumbent merchant silicon but with the in-house networking programs of the largest AI platform owners.

What happens next?

The $145 million gives CScale multi-year runway to move from development toward qualified product, though the company has not disclosed milestones tied to the capital. Whether it can win socket positions against entrenched optical and switch silicon vendors will determine whether this round becomes a foundation or a bridge.