Bringing the
speed of light to AI

Lumilens provides next-gen photonic interconnects for massively scaling out – and scaling up – AI compute infrastructure.

Learn More

LLMs & model
sizes exploding

Need for larger GPU clusters

Scale out &
scale up pressure

Physical limits
exceeded

Electrical + thermal
limits exceeded

This creates pressure on scale out architectures today, and how to design scale up architectures moving forward

Today’s massive AI data centers, and their leaders, need:

Faster innovation

Faster photonic innovation to keep up with AI workload growth – scale out and up

GPU direct connect

The ability to directly connect much larger sets of GPUs

Smarter manufacturing

Automated, data-centric manufacturing to maximize product yield + quality

Systems expertise

Suppliers with systems-level expertise to partner on scale up architectures

In-house developed silicon photonics

Supply chain continuity and more scalable manufacturing

Scale out + scale up

Only photonics vendor that enables cluster-to-cluster and direct GPU-to-GPU architectures

Much higher density GPU-to-GPU

Deploy 100x higher quantity of GPU direct connections

Scale Out

A faster pace of innovation, fully aligned with industry standards

Very high product quality via automated manufacturing + robotics + AI

Predictable system deployment and automatic data collection for troubleshooting

Scale Up

Ultra high-performance, scalable to 100x of existing limits with minimal latency

GPU disaggregation to achieve massive scale and mitigate power issues

Easily deployable using standards-based interconnects

Our customers benefit from:

Faster innovations vs. traditional photonics suppliers

Higher production efficiency and product quality

Supply chain continuity via broad ecosystem and in-house silicon photonics

Ability to manufacture anywhere

Photonics are the answer for scaling AI infrastructure going forward. While traditional scale out designs still have a lot of growth left, there is a massive opportunity to build native GPU-to-GPU scale up architectures. Our goal at Lumilens is to be the leader in both.

Ankur Singla

CEO & founder

AI and its required infrastructure continues to grow faster than the photonics industry can deliver solutions for it. That’s why we founded Lumilens – to participate in that growth, while designing new ways to better connect and scale AI natively.

Samuel Liu

VP Products & founder

Since our founding days, Lumilens has approached innovation in photonics at a very global level. Our India/APAC teams play a key role in delivering round-the-clock innovation and highly efficient manufacturing and operations.

Ritesh Kapahi

VP/GM India & founder

You must approach innovation in photonics from a system-level perspective, not a component-level one. That’s why we’ve built Lumilens with deeply experienced team members from across the board of leading networking and optical vendors.

Dave Friedman

VP Operations & founder

The industry need is clear…how to scale AI infrastructure to meet the pace of this massive technology revolution. And the answer is also clear…photonics to scale and reinvent ultra high-speed connectivity across GPUs.

Ted Schmidt

CTO

Photonics vendors have grown the market, but haven’t fully leveraged the actual AI that their connectivity enables. Lumilens is designing its full-stack manufacturing from the ground up, using robotics and AI to the fullest extent.

Weich Fang

VP Manufacturing & Ops

Our in-house silicon innovation enables higher-density integration and more efficient fiber utilization, forming a unique foundation for both scale up and scale out architectures in modern AI data centers.

Harish Devanagondi

VP Engineering, Silicon

I’ve been at many innovative networking companies, from the early days of LANs, the Internet boom, and the start of cloud computing. The role of photonics in scaling AI is as big, or bigger than, any of these. And Lumilens will be at the heart of that.