Lumilens Emerges from Stealth with More Than $900 Million in Funding to Break AI’s Connectivity Constraint

Lumilens, the connectivity platform for AI infrastructure led by Ankur Singla, Ritesh Kapahi, Samuel Liu, Dave Friedman, Ted Schmidt, and the team, has emerged from stealth, having been founded to solve the growing networking bottlenecks limiting AI scale. In just two years after its founding, the company is commercially shipping its first product into production AI data centers under a multi-billion-dollar customer agreement, and announced more than $700 million in new funding co-led by Atreides Management, Bain Capital Ventures, Meritech, Seligman Ventures and Spark Capital, with participation from Addition, Alkeon, Harbourvest, J.P. Morgan Private Capital, Mayfield, MVP Ventures, Qualcomm Ventures, Peak XV, Redpoint Ventures, Seifdune and other top-tier investors. The financing values the company at $5.51 billion and brings its total capital raised to over $900 million.

The world’s largest technology companies are investing heavily in AI infrastructure, with Citi projecting total spending will exceed $2.8 trillion by 2029. The challenge is increasingly shifting from acquiring GPUs to connecting them. Modern AI models run across hundreds of thousands of processors that must operate as a single system, making the network linking those chips as important to performance as the chips themselves.

This challenge is concentrated in two critical parts of the AI data center: scale-out and scale-up fabrics.

In the scale-out fabric connecting GPU racks and rows, each additional processor increases demand for optical transceivers. A 400,000-GPU data center requires more than 2.4 million transceivers and over five million fiber strands, putting further pressure on an already constrained supply chain. McKinsey projects that 800G optical transceiver production could fall 40–60% short of demand through 2027, while 1.6T transceiver shortages of 30–40% could persist through 2029.

Inside the rack, the scale-up fabric faces a physical limit as GPUs connect directly over copper. At the data rates required by AI workloads, electrical signals travel only around 1.5 meters over copper, limiting tightly coupled systems to a single rack and hundreds of GPUs. As hyperscalers move toward clusters of thousands of GPUs, overcoming this constraint will require replacing copper connections with photonics.

Lumilens was founded to address these connectivity constraints as AI clusters scale. In early 2024, the founding team began developing solutions for challenges already emerging at leading hyperscalers. Two years later, Lumilens’ first scale-out product completed qualification and began shipping to a hyperscaler’s production AI data centers under a multibillion-dollar agreement.

The company designs and manufactures optical interconnects for both major AI data center networks: scale-up systems that directly connect GPUs within a compute system and scale-out fabrics that connect racks and clusters.

Ankur Singla, founder and CEO of Lumilens, said the constraint on AI infrastructure has shifted from acquiring GPUs to connecting them at scale. Hyperscalers increasingly require greater optical capacity for existing networks, alongside technologies capable of directly connecting thousands of GPUs within a single cluster. Lumilens is designed to address both requirements while manufacturing at the volume, quality, and speed required for large-scale AI infrastructure.

Lumilens’ Connectivity Stack:

Lumilens’ portfolio spans both major connectivity layers in AI data centers:

• Scale-up silicon and systems (NPO and CPO): Near-package and co-packaged optics that bring optical I/O closer to GPUs, enabling thousands of processors, and eventually tens of thousands, to operate as a tightly coupled system.

• Scale-out pluggable transceivers (800G, 1.6T and beyond): Optical modules that replace copper links between racks and rows, delivering higher bandwidth at lower power for large-scale AI networks.

Every Lumilens product is built on LumiCore™, the company’s unified technology platform spanning silicon photonics, mixed-signal ICs, electrical-optical interposers and optical systems developed in-house. The shared platform enables Lumilens to move from design to customer-qualified products in months rather than years, while engineering each product for high-volume manufacturing from the outset.

Lumilens also treats manufacturing as a core product capability. Its electrical-optical interposer technology simplifies photonic chip assembly, supported by proprietary manufacturing processes, custom robotics, and automated testing. By combining manufacturing partners with its own large-scale facilities, the company aims to scale production alongside advances in optical performance and density.

Umesh Padval, Managing Partner at Seligman Ventures, said AI infrastructure constraints are increasingly shifting from access to compute toward connectivity and the ability to operate thousands of GPUs as a single system. He highlighted Lumilens’ team across silicon, optics, software, systems, supply chain and manufacturing, as well as its early hyperscaler adoption and multibillion-dollar customer commitments just two years after founding, as indicators of the scale of the opportunity and the company’s execution.

Gavin Baker, Managing Partner at Atreides Management, said AI infrastructure constraints are increasingly shifting from compute to connectivity. He highlighted Lumilens’ ability to move from founding to a qualified product deployed in hyperscaler production data centers within two years, while building manufacturing capacity to meet demand and diversify AI supply chains. He described the combination of speed and scale as a key strength behind the investment.

Navin Chaddha, Managing Partner at Mayfield, said GPU connectivity has become a key bottleneck as copper reaches its practical limits. He highlighted Lumilens’ focus on optical interconnects across both scale-up and scale-out AI infrastructure, alongside its approach of treating manufacturing as a core product capability. Mayfield has backed founder Ankur Singla since the company’s inception, leading its seed round and participating in subsequent financings. Chaddha noted Lumilens’ rapid growth to a valuation above $5 billion and multibillion-dollar customer orders within two years of founding.

James Kuklinski, General Partner at Spark Capital, said Lumilens is addressing AI connectivity challenges across both scale-up and scale-out networks. He highlighted the team’s expertise in silicon photonics, networking systems and high-volume manufacturing, as well as its rapid execution and market opportunity. Spark Capital first backed Lumilens by leading its Series B in early 2025 and later co-led its Series C financing.

Lumilens has assembled a leadership team with expertise across photonics and large-scale networking systems, including executives and engineers from Cisco, Juniper Networks, Meta, Marvell, Lumentum and Coherent, to pursue the more than $100 billion market opportunity for photonic interconnects.

The Lumilens Series C funding round includes participation from a wide range of top-tier investors: Addition, Aiconic, Alkeon, Atreides Capital, Bain Capital Ventures, EDBI (an arm of SG Growth Capital, the investment platform of EDB and Enterprise Singapore), Harbourvest, J.P. Morgan Private Capital, Lingotto, Mayfield, Meritech, MVP Ventures, Peak XV, Qualcomm Ventures, Redpoint Ventures, Seifdune, Seligman Ventures, Spark Capital and Thomvest Ventures. The new funding will help the Company expand its rapidly growing silicon, systems, software, process engineering, and high-volume manufacturing operations.

𝐑𝐞𝐦𝐨𝐭𝐞 𝐇𝐢𝐫𝐞 𝐖𝐢𝐭𝐡 𝐔𝐬: 𝐌𝐞𝐧𝐥𝐨 𝐓𝐚𝐥𝐞𝐧𝐭 helps technology companies hire exceptional remote talent from India. Build your team with carefully curated engineers, product leaders, designers, GTM professionals, and more. Learn More At: https://www.menlotimes.com/menlo-talent