Lumilens Exits Stealth at $5.5B Valuation to Fix AI Data Centers' Connectivity Crisis

Lumilens Emerges from Stealth with $5.51 Billion Valuation

Optical networking startup Lumilens exited stealth mode with a significant announcement: more than $700 million raised in a Series C funding round, bringing total capital to over $900 million and valuing the San Jose-based company at $5.51 billion1

1

2

2

. The round was co-led by Atreides Management, Bain Capital Ventures, Meritech, Seligman Ventures and Spark Capital, with participation from Qualcomm Ventures and JPMorgan among more than a dozen investors3

3

. Mayfield's Navin Chaddha, who has backed founder Ankur Singla three times, stated he had never witnessed growth like this in the firm's 56-year history2

2

.

Solving AI's Wiring Problem Through Optical Innovation

The two-year-old company addresses a fundamental shift in AI infrastructure constraints. "The constraint on AI has shifted from how many GPUs you can buy to how many you can connect," explained Ankur Singla, Lumilens CEO1

1

. Modern AI models run across hundreds of thousands of chips that must function as a single computer, making GPU connectivity the new bottleneck2

2

. Hyperscalers told Lumilens they need two critical capabilities: significantly more optical capacity for current networks and a path to directly connect thousands of GPUs into a single cluster1

1

.

Why Copper Hit a Wall in AI Data Centers

The AI industry faces two interconnected challenges that Lumilens targets. Across racks, every GPU added multiplies the optical transceivers required—a single 400,000-GPU site demands more than 2.4 million transceivers, exceeding market capacity2

2

. Inside racks, copper interconnects have reached their physical limits. At AI speeds, electrical signals survive approximately 1.5 meters over copper, capping tightly linked clusters at a few hundred chips2

2

. "The distance limitation starts to come in the moment you get to 1.6 Tb/s, you're actually at one meter or so in distance with copper. That basically is within the rack," Singla explained3

3

. To wire thousands of GPUs together, the industry must replace copper with light-based solutions.

LumiCore Architecture Powers Scale-Up and Scale-Out Networks

Lumilens designs and manufactures optical and networking products that enable AI data centers to move data faster while consuming less power1

1

. The company's technology addresses both scale-up networking, which links chips within a single rack, and scale-out networking, which connects different racks3

3

. For scale-up applications, Lumilens offers two product families: co-packaged optics (CPO) chips that integrate directly into GPUs for maximum power-efficient performance, and near-packaged optics (NPO) chips that sit on the motherboard hosting GPUs3

3

. "We believe that we're still about two to three years away when it comes to large-scale volume deployment of co-packaged optics and the interim solution is near-packaged optics," Singla noted, adding that Lumilens pioneered this approach with its first chip taping out in 20243

3

.

For scale-out transceivers, Lumilens ships silicon as part of pluggable devices attachable to switches, supporting connections at 800 gigabits per second, 1.6 terabits per second, or higher bandwidths without the range limitations plaguing copper3

3

. All products are built on the LumiCore architecture, a common set of building blocks including photonic components, mixed signal integrated circuits, and interposers that reduce chip development time from years to months3

3

.

Multi-Billion-Dollar Agreement Signals Market Validation

Lumilens is already shipping products under a multi-billion-dollar agreement with an unnamed customer identified as one of the four hyperscalers—Amazon, Alphabet, Microsoft, or Meta1

1

. The company says it has received billions of dollars worth of chip orders, demonstrating market validation beyond typical startup milestones3

3

. This deployment into live data centers gives Lumilens what the market values most: a product already solving real-world problems2

2

. The funding arrives as hyperscalers spend billions expanding AI data centers to meet surging demand for artificial intelligence services, with Citigroup forecasting AI infrastructure outlays exceeding $2.8 trillion by 20292

2

.

Proven Leadership and Competitive Landscape

Ankur Singla brings credibility through his track record of building and selling two infrastructure companies: Contrail to Juniper and Volterra to F5, for a combined $676 million3

3

. Lumilens has assembled a team of executives and engineers from Cisco, Juniper Networks, Meta, Marvell, Lumentum and Coherent, bringing deep expertise in photonics and large-scale networking1

1

. However, the company faces intense competition. Co-packaged optics represents one of the hottest corners of the chip world, contested by Nvidia, Broadcom, and numerous startups2

2

. Chinese rivals including Eoptolink are pursuing the same prize, and while a substantial private valuation is notable, maintaining a durable lead requires continuous innovation2

2

.

What to Watch: Manufacturing Scale and Market Expansion

Lumilens plans to use the fresh capital to scale engineering and manufacturing operations1

1

. The company has developed a custom manufacturing workflow incorporating robots and AI software to ease chip assembly processes3

3

. With its second generation of chips taped out and preparing for commercial deployment next year, the company's ability to execute at volume will determine whether it can maintain its early advantage3

3

. The broader implication is clear: in high-speed optical networking for AI, moving data has become harder than processing it, and the companies that solve this wiring problem will shape the future of AI infrastructure.