Lightmatter

AI Hardware

Lightmatter is a photonic computing company founded in 2017 by MIT researchers Nicholas Harris, Thomas Graham, and Darius Bunandar.

Reviewed by 7wData
Share Email

On this page

Profile

Builds photonic semiconductor interconnects that use light instead of electricity to move data between chips in AI data centers, eliminating bandwidth bottlenecks.

Lightmatter is a photonic computing company founded in 2017 by MIT researchers Nicholas Harris, Thomas Graham, and Darius Bunandar. The company develops silicon photonics technology to solve bandwidth constraints in large-scale AI infrastructure by using light instead of electricity to move data between chips. Harris serves as co-founder and CEO.\n\nThe company's product portfolio centers on the Passage platform—a suite of photonic interconnects ranging from co-packaged optics to 3D photonic interposers.

The flagship offerings include the L200 (32-64 Tbps), L20 (12.8 Tbps), and M1000 superchip (114 Tbps total optical bandwidth). Lightmatter also offers the Envise photonic processor for AI compute acceleration and the Guide light engine, a Very Large Scale Photonics (VLSP) module enabling unprecedented laser bandwidth.\n\nLightmatter has raised $850 million across five funding rounds, culminating in a $400 million Series D in October 2024 that valued the company at $4.4 billion—a near 3.7x increase from its December 2023 valuation of $1.2 billion. The October 2024 round was led by T.

Rowe Price and included participation from existing investors Fidelity and Google Ventures. CEO Harris stated the round is "probably our last private funding round," signaling IPO preparation. In July 2024, the company appointed Simona Jankowski, former Nvidia VP of investor relations and strategic finance, as CFO—a move underscoring IPO planning.\n\nThe company targets hyperscalers training frontier AI models with 100,000+ GPU clusters (Microsoft, Amazon, Google, Meta, OpenAI, xAI).

Lightmatter does not disclose specific customers publicly, positioning itself as a foundry-like supplier. Manufacturing partnerships with GlobalFoundries, ASE, and Amkor began with late-2025 delivery targets. In January 2026, Lightmatter announced a strategic partnership with Global Unichip Corp. (GUC) for co-packaged optics integration and advanced packaging.\n\nRecent technical milestones in 2026 include a record 1.6 Tbps per-fiber achievement (March 2026) and the appointment of Roy Kim as VP of Product (April 2026), signaling acceleration toward commercial deployment. Glassdoor reviews present a mixed picture: employees praise the company's innovation focus and technical caliber but cite concerns about leadership accountability and work-life balance consistency.

Track Lightmatter and 240+ vendors.

335k+ subscribers read the daily AI & data note. One email, both newsletters. Unsubscribe anytime.

Who buys this

  • Hyperscalers training frontier AI models (100,000+ GPU clusters)
  • Large-scale AI infrastructure operators (cloud providers, research labs)
  • Data center equipment manufacturers integrating high-bandwidth optical systems
  • GPU and accelerator chip manufacturers requiring advanced interconnect solutions
  • AI model training facilities with trillion-parameter model requirements

Strengths and what to watch

Strengths

  • Only company shipping 3D co-packaged optics at scale for AI; L200 and M1000 architectures achieve 5-10x bandwidth improvement over NPO solutions.
  • Secured $400M Series D at $4.4B valuation with backing from T. Rowe Price, Google Ventures, and Fidelity; CEO signals probable final private round before IPO.
  • Leadership team combines deep photonics PhDs (Harris: 87+ patents, Optica Fellow, MIT-trained) with CFO from Nvidia's investor relations and VP of Product from industry operations.

Watch for

  • No named customers disclosed; early-stage production at scale (late 2025 target). Manufacturing execution via GlobalFoundries, ASE, Amkor critical to timelines.
  • Photonic systems at scale remain historically challenging; silicon photonics yield and thermal management unproven at hyperscaler volumes.
  • Mixed Glassdoor reviews flag leadership accountability and work-life balance concerns; rapid scaling from 2024–2026 may trigger turnover.

Recent moves

Key Information

Industry
AI Hardware
Founded
2017
Headquarters
United States

Frequently Asked Questions

What does Lightmatter do?

Lightmatter builds photonic semiconductors using light instead of electricity to move data between AI chips. Founded in 2017 by MIT researchers, the company's Passage platform includes co-packaged optics and 3D photonic interposers, solving bandwidth bottlenecks in large-scale AI infrastructure and data centers.

How much has Lightmatter raised in funding?

Lightmatter has raised $850 million across five funding rounds. In October 2024, the company closed a $400 million Series D led by T. Rowe Price, valuing Lightmatter at $4.4 billion—a 3.7x increase from its $1.2 billion December 2023 valuation.

What products does Lightmatter offer?

Lightmatter's Passage platform includes the L200 (32-64 Tbps), L20 (12.8 Tbps), and M1000 superchip (114 Tbps optical bandwidth). The company also manufactures Envise, a photonic processor for AI compute acceleration, and Guide, a light engine module enabling unprecedented laser bandwidth capabilities.

Is Lightmatter shipping products to customers yet?

Lightmatter does not disclose specific customers publicly, positioning itself as a foundry-like supplier targeting hyperscalers training frontier AI models. Manufacturing partnerships with GlobalFoundries, ASE, and Amkor began with late-2025 delivery targets. The company is in early-stage production scaling as of 2026.

Is Lightmatter planning an IPO?

CEO Nicholas Harris stated the October 2024 Series D round is 'probably our last private funding round,' signaling IPO preparation. The company appointed Simona Jankowski, former Nvidia VP of investor relations and strategic finance, as CFO in July 2024—further underscoring IPO readiness.

What recent technical milestones has Lightmatter achieved?

In March 2026, Lightmatter achieved a record 1.6 Tbps per-fiber throughput in co-packaged optics using 16-wavelength DWDM and 112G Qualcomm SerDes. In January 2026, the company announced a strategic partnership with Global Unichip Corp. for advanced integration, accelerating commercial deployment timelines.

How Lightmatter compares

Direct head-to-head against 3 competitors. Picked by 7wData.

This company

Lightmatter

Positioning
Builds photonic semiconductor interconnects that use light instead of electricity to move data between chips in AI data centers, eliminating bandwidth bottlenecks.
Customer segments
Hyperscalers training frontier AI models (100,000+ GPU clusters)
Strengths
Only company shipping 3D co-packaged optics at scale for AI; L200 and M1000 architectures achieve 5-10x bandwidth improvement over NPO solutions.
Watch for
No named customers disclosed; early-stage production at scale (late 2025 target). Manufacturing execution via GlobalFoundries, ASE, Amkor critical to timelines.
Recent moves
Lightmatter names Roy Kim as VP of Product to lead global deployment of photonic interconnects

Ayar Labs

Positioning
In-package optical I/O chiplets for AI accelerators, co-developed with chip manufacturers targeting GPU bandwidth constraints.
Customer segments
GPU chipmakers, AI accelerator designers, hyperscalers building next-generation training clusters. Buyers in Nvidia and AMD supply chains.
Strengths
TeraPHY chiplet co-developed with Nvidia and AMD as both investors and co-design partners, embedded directly in chip design cycles.
Watch for
Volume production roadmap targets 2028. No confirmed hyperscaler deployment as of mid-2026 despite $870M raised.
Recent moves
$500M Series E closed March 2026, backed by Nvidia, AMD, and MediaTek. Valuation $3.75B, total raised $870M.

Celestial AI (Marvell)

Positioning
Photonic Fabric platform for in-die optical delivery, absorbed into Marvell's semiconductor portfolio via acquisition December 2025.
Customer segments
Hyperscalers and AI ASIC designers needing memory bandwidth disaggregation. Buyers already sourcing Marvell networking silicon.
Strengths
Claims 25x bandwidth over standard CPO by delivering optical signals to any die location, not only chip edges.
Watch for
Marvell acquisition introduces roadmap absorption risk. Celestial technology not expected to generate meaningful revenue until late 2028.
Recent moves
Marvell announced $3.25B acquisition of Celestial AI in December 2025, targeting close by end of Q1 2026.

Ranovus

Positioning
Monolithic electro-photonic IC platform for co-packaged optics, targeting AI, HPC, and high-bandwidth switching applications.
Customer segments
AI compute integrators, HPC cluster builders, and switch vendors requiring terabit-scale optical I/O in compact form factors.
Strengths
ODIN EPIC integrates laser, modulator, photodetector, and driver on a single monolithic die, reducing co-packaged assembly complexity.
Watch for
No publicly disclosed funding rounds or valuation. Financial scale and production capacity uncertain against better-capitalized rivals.
Recent moves
March 2026 partnership with Cerebras Systems targeting wafer-scale co-packaged optics for AI supercomputing infrastructure.

Sources

  1. lightmatter.co — Company mission, product portfolio (Passage, Envise, Guide), and technical specifications
  2. techcrunch.com — Series D funding details ($400M, $4.4B valuation), CEO statement on final private round, market positioning, and target customers
  3. news.crunchbase.com — Complete funding history, valuation trajectory from $1.2B (Dec 2023) to $4.4B (Oct 2024), and competitive positioning in photonics space
  4. www.crunchbase.com — Founder names, founding year (2017), total funding raised ($850M), and funding round details across Series A through D
  5. www.businesswire.com — CFO appointment of Simona Jankowski from Nvidia, her background in investor relations and strategic finance
  6. www.glassdoor.com — Employee culture reviews, ratings (4/5 stars, 79% would recommend), and mixed feedback on leadership and work-life balance