🔥Stalecollected in 5m

Nvidia-Backed Ayar Labs Raises $500M at $3.75B

Nvidia-Backed Ayar Labs Raises $500M at $3.75B
PostLinkedIn
🔥Read original on 36氪

💡Nvidia-backed optical chips firm valued at $3.75B—vital for AI data center bandwidth crunch.

⚡ 30-Second TL;DR

What Changed

Ayar Labs secures $500M Series E funding

Why It Matters

This funding accelerates optical interconnect development critical for AI data centers needing ultra-high bandwidth. It signals strong investor confidence in photonics for scaling AI infrastructure amid GPU bottlenecks.

What To Do Next

Evaluate Ayar Labs optical I/O for your next AI cluster interconnect upgrade.

Who should care:Enterprise & Security Teams

🧠 Deep Insight

Web-grounded analysis with 9 cited sources.

🔑 Enhanced Key Takeaways

  • Ayar Labs' TeraPHY optical I/O chiplet achieves 8 Tbps bandwidth using UCIe standard and microring resonator-based optical engine with SuperNova 16-wavelength external light source[2][6][7].
  • Company founded in 2015 by MIT alumni Mark Wade, Alex Wright, and Chen Sun, who won 2015 MIT Clean Energy Prize for Optibit prototype enabling world's first optical microprocessor-to-memory communication[3][5].
  • Strategic partnerships include HPE for supercomputing, Intel for FPGA demos, GlobalFoundries for 45nm photonics manufacturing, and recent collaboration with Alchip for AI infrastructure scaling[2][5][9].
  • Holds over 200 patents in photonics; demonstrated 2Tbps over 8 single-mode fibers using 32Gbps NRZ per channel, more efficient and lower latency than PAM4[2][6].

🛠️ Technical Deep Dive

  • TeraPHY chiplet: 8 Tbps bandwidth via UCIe optical chiplet, microring resonator optical engine, supports co-packaged optics (CPO) for AI scale-up[6].
  • SuperNova light source: Second-gen, CW-WDM MSA-compliant 16-wavelength external laser, drives 256 optical carriers across 16 single-mode fibers, outperforms VCSEL AOCs and DR4/DR8[6][7].
  • 2Tbps demo: 8 channels of 32Gbps NRZ per single-mode fiber, lower latency without FEC, excludes external laser[2].
  • Process nodes: 45nm silicon photonics on GlobalFoundries 45CLO, first working silicon in 2020[2][5].
  • Optical I/O: In-package chiplets replace electrical I/O, higher bandwidth, lower power/latency/reach than electrical or pluggable optics[4].

🔮 Future ImplicationsAI analysis grounded in cited sources

Ayar Labs' optical I/O will enable disaggregated memory clusters for GPUs, reducing AI memory wall bottlenecks
SuperNova and TeraPHY provide high-bandwidth, low-latency optical links essential for scaling AI workloads beyond electrical limits[7].
Co-packaged optics adoption will accelerate with Nvidia/HPE partnerships for next-gen HPC architectures
Strategic investments and demos position Ayar as leader in CPO for GPU fabrics and supercomputers[2][3].
Optical interconnects will cut AI data center power by replacing inefficient electrical links
Higher efficiency and density of silicon photonics solutions address bandwidth-distance bottlenecks in AI infrastructure[1][4].

Timeline

2015-05
Founded by MIT alumni after winning Clean Energy Prize for optical prototype
2015-10
$2M seed round; prototype 45nm optical chip
2020-04
$35M Series B; first 45nm silicon photonics silicon
2022-06
$130M Series C; first volume shipments, partnerships with HPE/NVIDIA/Lockheed
2024-12
$155M Series D led by Advent/Light Street with AMD/Intel/NVIDIA; unicorn valuation >$1B
2025-03
Opened Taiwan office; UCIe optical chiplet and GTC CPO announcements
📰

Weekly AI Recap

Read this week's curated digest of top AI events →

👉Related Updates

AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪