Lightmatter company profile: valuation, funding & revenue

This Lightmatter company profile covers the company's valuation, revenue, funding rounds, investors and history. Lightmatter works in semiconductors, with a focus on photonic interconnect and computing for AI. It was founded in 2017 and is headquartered in Mountain View, California, United States. Its latest valuation on record is US$4.4 billion (October 2024). Every figure links to a public source, and anything the company hasn't disclosed is marked as such.

Revenue

Funding history

Lightmatter timeline

Key dates from the sources on this page, oldest first. Metric snapshots are listed after the funding events.

• 2017 — Lightmatter is founded in Mountain View by Nick Harris (CEO), Darius Bunandar and Thomas Graham.
• Dec 2023 — Series C extension: US$155M, valuing the company at US$1.2B.
• Oct 2024 — Series D: US$400M, valuing the company at US$4.4B. Investors: T. Rowe Price-advised accounts (lead); GV, Fidelity.
• October 2024 — Total funding: About US$850 million.

Investors

Overview

Lightmatter develops photonic technology, using light instead of electrical signals, to move data between AI chips faster and with far less energy. Its main product, Passage, is a photonic interconnect: a platform that lets GPUs and other processors communicate through optical connections built directly into the chip package. As AI models are trained across tens of thousands of chips, the speed and power cost of moving data between them has become as important a bottleneck as raw computing power. Lightmatter aims to remove that bottleneck, enabling larger and more efficient AI supercomputers.

Nick Harris, the CEO, Darius Bunandar and Thomas Graham founded Lightmatter in 2017 after research at the Massachusetts Institute of Technology, where Harris and Bunandar worked on photonic computing. The company began in Boston and is now headquartered in Mountain View, California, with engineering in both regions.

Why photonics matters for AI

Inside a data centre, AI chips constantly exchange data: model weights, intermediate results and gradients during training. Traditional copper connections struggle with distance and consume increasing amounts of power at higher speeds, while conventional optical transceivers, which convert electrical signals to light at the edge of servers, add cost, latency and energy. Building optical connections directly alongside chips, often called co-packaged optics, can move far more data per second with less power. That makes it possible to connect more chips together as if they were one giant processor, which is exactly what training frontier models requires.

Products

• Passage: a photonic interconnect platform that provides high-bandwidth optical connections between chips, offered in versions designed for different packaging approaches.
• Photonic computing research: Lightmatter's earlier Envise product explored performing AI calculations using light itself; the company has since focused commercially on interconnect, where demand is more immediate.
• Partnerships: collaborations with chip designers, foundries and packaging companies to integrate Passage into next-generation AI systems.

Funding

Lightmatter raised from investors including GV (Google's venture arm), Spark Capital, Matrix Partners, Viking Global Investors and HPE Pathfinder. A US$155 million extension of its Series C in December 2023 valued it at US$1.2 billion. In October 2024 it raised a US$400 million Series D at a US$4.4 billion valuation, led by accounts advised by T. Rowe Price, with GV and Fidelity participating. Bloomberg noted the valuation almost quadrupled from the prior round. Total funding reached about US$850 million. We haven't found a newer priced round.

Business model

Lightmatter sells and licenses photonic interconnect technology to companies building AI processors and systems, including chip designers and hyperscale data centre operators. Revenue may come from product sales, licensing and engineering partnerships. The company hasn't disclosed revenue, so we show it as not disclosed. Its success depends on being designed into future generations of AI hardware, which involves long development cycles.

Competition

Lightmatter competes with other silicon photonics companies such as Ayar Labs and Celestial AI (acquired by Marvell), and with large incumbents developing co-packaged optics, including Nvidia, Broadcom, Intel, Cisco and Marvell. Its advantages are an early lead in integrating photonics directly with chips, deep research expertise and strong funding. Our guide to AI chip startups covers related companies such as Etched and Tenstorrent.

Lightmatter at a glance

Lightmatter is a Mountain View photonics company founded in 2017 by MIT researchers Nick Harris, Darius Bunandar and Thomas Graham. Its Passage platform connects AI chips with light to increase bandwidth and cut energy use. It raised a US$400 million Series D at a US$4.4 billion valuation in October 2024, led by T. Rowe Price, bringing total funding to about US$850 million.

Timeline

• 2017: founded by MIT photonics researchers.
• 2021: raises a Series B and develops photonic computing products.
• December 2023: extends its Series C to US$155 million at US$1.2 billion.
• October 2024: raises US$400 million at US$4.4 billion, led by T. Rowe Price.
• 2025: announces new Passage products aimed at next-generation AI systems.

The energy problem in AI data centres

AI data centres consume enormous amounts of electricity, and power availability now limits how fast new capacity can be built. A significant share of that energy goes not into computation but into moving data around. Technologies that reduce the energy cost of communication can therefore increase the useful computing a data centre can deliver within its power budget. This makes photonics part of the broader effort to make AI more energy-efficient, alongside better chips, cooling and power supply.

How co-packaged optics changes system design

Today, AI servers typically connect chips using electrical links within a server and pluggable optical modules between servers. Co-packaged optics moves the optical components into the same package as the processor, shortening electrical paths and allowing much higher bandwidth. This could let system designers build "scale-up" domains with many more chips acting together. Adoption requires changes across the supply chain, from chip packaging to data centre maintenance, so it is expected to roll out gradually over several hardware generations.

Risks

• Design wins: success depends on being chosen for major AI chip platforms.
• Incumbent competition: Nvidia, Broadcom and others are investing heavily in their own optics.
• Manufacturing: photonic integration requires complex, specialised processes.
• Timing: adoption of co-packaged optics could be slower than expected.

Questions for system designers

• What bandwidth and energy savings does the interconnect deliver in real systems?
• How does it fit with your packaging and manufacturing partners?
• How are reliability and serviceability handled in data centres?

Who buys photonic interconnect

The main customers for technology like Passage are companies designing AI accelerators and the hyperscale cloud providers and AI labs that build the largest training clusters. These buyers plan hardware generations years in advance, so photonics suppliers work closely with them during design. Network equipment makers and server manufacturers are also potential partners, because they integrate optical components into the systems sold to data centres.

Latest developments (October 2026)

Lightmatter has not announced a new round since its October 2024 Series D at US$4.4 billion. In January 2026 it partnered with Global Unichip Corp. to commercialise its Passage 3D co-packaged optics, and in March it became a founding member of an industry agreement on optical interconnects, according to Converge Digest. Photonic interconnects are becoming important for linking large numbers of AI chips. See our AI chip startups guide.

What to watch

Watch for announced customer design wins, volume production, new funding or IPO plans, and progress by competitors in co-packaged optics.

Takeaways

AI infrastructure builders may use photonics to scale clusters efficiently. Founders can see how university research becomes infrastructure for a booming market. Investors should watch for concrete design wins to support the valuation.

Sources

Every figure on this page comes from the public sources below. Where sources disagreed, we used the company's own announcement or a primary news report. Private-company data changes; check the date beside each figure.

1. BNN Bloomberg
2. DatacenterDynamics