Lumilens, a San Jose-based optical interconnect startup founded in early 2024, emerged from stealth with a Series C round of more than $700 million that values the company at $5.51 billion. Total capital raised now exceeds $900 million.
Lumilens’ $700 million financing was co-led by Atreides Management, Bain Capital Ventures, Meritech, Seligman Ventures, and Spark Capital. Participating investors include Addition, Alkeon, HarbourVest, J.P. Morgan Private Capital, Mayfield, MVP Ventures, Qualcomm Ventures, Peak XV, Redpoint Ventures, Seifdune, and others such as EDBI, Lingotto, and Thomvest.
Proceeds will expand engineering (silicon, systems, software, process) and high volume manufacturing operations, including custom robotics, test automation, and proprietary process recipes.
What is Lumilens?
Lumilens designs and manufactures photonic interconnects specifically for AI data centers. Its core platform, LumiCore, integrates in-house silicon photonics, mixed signal ICs, electrical-optical interposers, and optical systems. This single platform underpins both scale-out and scale-up products, enabling faster design to qualification cycles and high volume manufacturing from the start.
The company addresses two primary constraints in large AI clusters:
- Scale-out: Pluggable 800G and 1.6T (and beyond) optical transceivers that link racks and rows, replacing or augmenting copper while delivering higher bandwidth at lower power.
- Scale-up: Near package optics (NPO) and co-packaged optics (CPO) that bring optical I/O directly to or near GPUs, extending tightly coupled domains beyond the ~1.5 meter copper limit to support thousands (and eventually tens of thousands) of GPUs acting as a single system.
Founder and CEO Ankur Singla is a serial infrastructure entrepreneur. Previous companies include Contrail Systems (acquired by Juniper Networks) and Volterra (acquired by F5). He has also held senior roles including CTO at Aruba. CTO and co-founder Ted Schmidt previously advanced silicon photonics and optical integration at Juniper (building on its Aurrion acquisition) and held senior photonics roles at Lumentum. The broader team includes executives and engineers from Cisco, Juniper, Meta, Marvell, Lumentum, Coherent, and Nokia, spanning photonics, mixed signal design, systems, manufacturing, and hyperscaler scale networking.
Lumilens treats manufacturing itself as a product: proprietary interposer technology, robotics, AI optimized processes, test automation, and MES systems designed for hyperscaler volumes, yield, and supply chain resilience. It operates with teams in the U.S. and India/APAC for continuous development and efficient production.
Within roughly two years of founding, Lumilens completed qualification of its initial scale-out product and began commercial shipments into production AI data centers of one of the four major hyperscalers (Amazon, Google, Microsoft, or Meta; specific name undisclosed). The company describes the customer agreement as multi billion dollars in value over the coming years.
This early production deployment at hyperscaler scale, combined with the rapid capital raise, is frequently cited by investors as exceptional execution for a hardware company of this age and complexity.
AI infrastructure spending is projected by Citigroup to exceed $2.8 trillion by 2029. The primary constraint has shifted from GPU availability to the ability to interconnect them effectively so that hundreds of thousands of processors function as a single coherent system.
Scale-out demand is extreme: a 400,000 GPU data center can require more than 2.4 million optical transceivers and over five million fiber strands. McKinsey has projected significant shortfalls, 40–60% for 800G transceivers through 2027 and 30–40% for 1.6T through 2029.
Recommended: American Growth Insurance Raises $70 Million In Funding
Scale-up faces hard physical limits with copper at AI data rates, restricting tightly coupled domains to roughly a single rack. Optical solutions (NPO/CPO) are required to scale further while managing latency, power, and density. Lumilens positions itself as one of the few suppliers covering both layers with a unified, manufacturable platform.
Investor commentary emphasizes the combination of proven founder execution, early hyperscaler design wins and revenue, in-house silicon and manufacturing control, and timing relative to the AI networking bottleneck. Mayfield’s Navin Chaddha (who has backed Singla multiple times) described the two year trajectory to a multi billion dollar valuation and multi billion dollar customer commitments as unprecedented in the firm’s history.
The new capital is directed at scaling silicon design, systems and software engineering, process development, and high volume manufacturing capacity to meet hyperscaler demand and diversify supply chains in a capacity constrained optics market. The company already claims the ability to manufacture at true hyperscaler volumes through a combination of proprietary automation and partner facilities.
Lumilens enters a competitive landscape that includes established optical module and photonics suppliers as well as other startups and larger players pursuing co-packaged and near package optics (with Nvidia and others also advancing photonics roadmaps). Its differentiation rests on the integrated LumiCore platform, dual coverage of scale-up and scale-out, manufacturing first design philosophy, and demonstrated early production shipment under a large hyperscaler contract.
At a $5.51 billion valuation after only two years and with commercial shipments underway, the round reflects strong investor conviction in both the structural demand for high performance, power efficient optical interconnects in AI data centers and Lumilens’ ability to execute at the required technical and production scale.
Please email us your feedback and news tips at hello(at)superbcrew.com
