On March 3, 2026, San Jose-based Ayar Labs announced a $500 million Series E at a $3.75 billion post-money valuation, bringing total funding to roughly $870 million. The company exists for one purpose: replacing copper with light inside AI data centers. Crunchbase’s weekly tally of the biggest U.S. rounds (February 28 to March 6) put Ayar Labs alongside Sierra — a signal that while model launches soak up the headlines, capital is starting to price in a much more physical bottleneck.
Why does this matter? CEO Mark Wade said it most directly: “AI infrastructure is hitting a power wall driven by interconnect inefficiency.” GPUs keep getting faster, but the kilometers of copper stitching thousands of them into one cluster have become a simultaneous limit on bandwidth and power. This round is a bet on how quickly that wall comes down.
A $500M Series E: Who Put In the Money
Neuberger Berman led the round and takes a board observer seat. New investors include Alchip Technologies, MediaTek, ARK Invest, Insight Partners, the Qatar Investment Authority, Sequoia Global Equities, and 1789 Capital. Existing financial investors Advent Global Opportunities, Boardman Bay Capital Management, IAG Capital Partners, Light Street Capital, and Playground Global all re-upped.
The strategic names are the more interesting read: AMD Ventures and NVIDIA both reinvested. The two GPU leaders putting money into a company whose product replaces copper is an admission that interconnect sits on the critical path before they can sell many more accelerators. It also extends a pattern — Ayar Labs’ $155 million Series D in 2024 already counted AMD, Intel Capital, and NVIDIA among its backers.
The use of funds is concrete: scale high-volume production and test capacity for its co-packaged optics (CPO) solution, expand globally (including a new office in Hsinchu, Taiwan), deepen ecosystem partnerships, and accelerate deployment.
Why Interconnect: The Power Wall Behind AI Clusters
The math of cluster scaling is unforgiving. Bigger models demand that scale-up networks deliver more bandwidth per GPU, and every step up in electrical bandwidth costs more power and reaches less far. Copper has a physical ceiling — signal integrity, reach, and energy per bit all tighten at once.
Wade’s framing compresses the argument into one line: CPO lets “thousands of GPUs operate as a unified system.” Gabe Cahill, managing director at Neuberger Berman, supplied the capital-markets version: the AI buildout is “one of the largest capital deployment opportunities of our generation,” and “data center interconnect has quickly emerged as the most critical bottleneck.” However fast models improve, the physical wiring inside the building decides whether all that compute actually fuses together.
TeraPHY: An Optical Engine Inside the Accelerator Package
Ayar Labs’ core product is the TeraPHY optical engine — HPCwire’s coverage describes 8 Tbps-class chiplets — paired with the SuperNova light source to form a complete optical I/O solution. The decisive word is “co-packaged”: instead of living in a rack-scale switch, the optical engine sits in the same package as the accelerator or switch silicon, converting electrical signals to light at the shortest possible distance.
The engineering is deliberately unglamorous. TeraPHY is built on the standard form factors, manufacturing, and packaging flows that major accelerator and switch vendors already use, which lowers adoption friction: CPO can ship without waiting for the whole packaging ecosystem to be reinvented. Note who came in as new money — Alchip and MediaTek are ASIC-ecosystem players, and their investment doubles as pre-integration work; the Hsinchu office puts engineering staff in the middle of the Asian supply chain.
What It Means for Buyers and Architecture Teams
Three practical effects. First, optical interconnect has moved from research topic to production schedule: the round’s stated purpose is scaling volume production and test capacity, which makes 2026 the year CPO goes from samples to deployments — teams designing next-generation racks should put optical options into their power and bandwidth models now. Second, the strategic investor roster is a weathervane: with both NVIDIA and AMD backing an interconnect startup, the composability of accelerator roadmaps with optical solutions will shape next-generation cluster procurement. Third, for cloud and sovereign-AI builders, interconnect efficiency is electricity and rack density — with power now the hard limit on data center expansion, this is an architectural constraint that shows up earlier than model choice does.
Sources
- Ayar Labs Closes $500M Series E, Accelerates Volume Production of Co-Packaged Optics — Ayar Labs
- Ayar Labs Closes $500M Series E — HPCwire
- The Week’s 10 Biggest Funding Rounds — Crunchbase News
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.