AI startup Lumilens has emerged from stealth with more than $900 million in total funding after closing a Series C round exceeding $700 million. The company, valued at $5.51 billion, develops photonic interconnect technology designed to connect thousands of GPUs as a single AI system inside data centers.
The funding underscores a growing industry consensus that AI’s scaling bottleneck has shifted from individual chip performance to cluster-level connectivity. Lumilens estimates that a single data center containing 400,000 GPUs could require more than 2.4 million optical transceivers and over five million fiber strands, a demand that current copper-based electrical connections cannot efficiently meet.
Copper interconnects still dominate inside AI racks, but at higher data rates they create reach limitations, power overhead, and thermal constraints. Lumilens argues that photonics must replace copper for both short hops within racks and longer connections between them as models scale beyond current cluster sizes.
The company’s platform, called LumiCore, combines silicon photonics, mixed-signal integrated circuits, electrical-optical interposers, and optical systems into a single product family covering pluggable transceivers, near-packaged optics, and co-packaged optics. Lumilens says it has completed qualification of its first product and begun shipping into a hyperscaler’s production data centers.
The startup was co-founded by former Intel executives and is based in San Jose. Its emergence adds pressure to a crowded optical interconnect market that includes established players like Coherent, Lumentum, and Broadcom, as well as Chinese rivals such as Eoptolink. The technology is critical infrastructure for both Nvidia GPU clusters and Google TPU systems.
Optical interconnect has become one of AI infrastructure’s hottest investment areas. Lumentum, whose stock has surged 340% since 2025 on demand for its laser and switching components, recently projected $8 billion in annualized revenue. The global market for optical interconnect in AI data centers is expected to grow rapidly through 2033 as hyperscalers race to deploy next-generation clusters.
Sources:
Lumilens Official Announcement
discussion