Thursday, 24 Sep 2026
Subscribe to AIWatcher
AIWatcher
  • Home
  • News

    OpenAI slots two cheaper GPT-6 models beneath Astra

    By
    AIWadmin

    Researcher hijacks Meta’s Muse agent through a hidden setting

    By
    AIWadmin

    ChatGPT learns to take voice orders for office chores

    By
    AIWadmin

    YouTube hands creators an AI agent for titles and thumbnails

    By
    AIWadmin

    Listeners can now rewrite the Spotify algorithm in plain words

    By
    AIWadmin

    Anthropic’s Claude agents flag a new enzyme family in viral DNA

    By
    AIWadmin
  • Articles

    Data shop Snorkel AI banks $350M as labs stockpile training sets

    By
    AIWadmin

    Ema banks $77M to push AI employees into the back office

    By
    AIWadmin

    US military command randomises routes to outfox enemy models

    By
    AIWadmin

    Kyutai teaches a speech model to do arithmetic out loud

    By
    AIWadmin

    Nokia hands developers a no-training route to calibrated answers

    By
    AIWadmin

    An open model sorts eight overlapping speakers in real time

    By
    AIWadmin
  • Spotlight

    DeepMind keeps cloud memory encrypted behind device-held keys

    By
    AIWadmin

    Anthropic trims Opus costs and speeds output in a mid-cycle refresh

    By
    AIWadmin

    Cisco Talos builds a fingerprint library for AI-driven malware

    By
    AIWadmin

    Google parks idle agents in a new open source runtime

    By
    AIWadmin

    OpenAI gives mathematicians a voice but hands them no brake

    By
    AIWadmin

    NVIDIA teaches its robotics stack to take instructions from agents

    By
    AIWadmin
  • About
    • Mission
    • Services
    • Contact
  • Newsletter
  • Shop
    • All Items
    • By Category
      • Hats
      • T-Shirts
    • Cart
  • 🔥
  • Alignment
  • Classification
  • Distillation
  • Explainability
  • Hallucination
  • Legal/Compliance
  • Medical
  • NLM
  • Mobility
  • Research
  • Robotics
  • Safety
  • Startups
  • Prompt
  • Python
  • RAG
  • RLHF
  • Token
  • Vision
Font ResizerAa
AIWatcherAIWatcher
  • Home
  • News
  • Articles
  • Spotlight
  • About
  • Newsletter
  • Shop
Search
  • Home
  • News
  • Articles
  • Spotlight
  • About
    • Mission
    • Services
    • Contact
  • Newsletter
  • Shop
    • All Items
    • By Category
    • Cart
Have an existing account? Sign In
Follow US
© 2022 Foxiz News Network. Ruby Design Company. All Rights Reserved.
News

Lightmatter halves fiber counts with a bidirectional optical module

Lightmatter has joined the Open CPX MSA and shipped a module that puts both directions of an optical link on one fiber.

AIWadmin
Last updated: September 18, 2026 1:17 am
AIWadmin
ByAIWadmin
Global AI news & information.
Follow:
Share
SHARE

Copper stops being useful at roughly one meter per lane, which is why an electrically wired accelerator cluster tops out at one or two racks.

Lightmatter is betting the next tier of scale-up networking has to be optical, and it has joined the Open CPX multi-source agreement to argue the point with hardware. Its new Passage L20 CPX module, which the company calls the first Open CPX module for AI infrastructure using bidirectional links, carries 6.4 Tbps in each direction for 12.8 Tbps total across 32 lanes per direction at a 212.5 Gbps PAM4 rate.

Bidirectional optics work by assigning each direction its own wavelength, so transmit and receive share a single strand instead of using two. Founder and chief executive Nick Harris said AI infrastructure is outgrowing copper, and that shipping optics and fiber into production now means the hard parts are already proven when co-packaged optics arrive.

The arithmetic at pod scale is the pitch. Lightmatter estimates a 512-GPU pod drops from about 131,000 fibers to 65,000, sheds roughly 16,000 connectors and more than 200 miles of fiber, and takes about 15 percent off total scale-up interconnect cost. Eliminating the shuffle boxes that route those connectors also halves the passive components that can fail.

A September 14 technical brief from Taylor Groves and Lars Johnsson supplies the detail: the conventional link architecture needs 32,768 connector endpoints cleaned and tested, work the authors put above 1,000 hours per pod.

The module runs on 1331 and 1311 nanometre O-band wavelengths, spans more than 500 meters on single-mode fiber, and pairs with Lightmatter’s Guide DR laser or compatible sources. Evaluation kits are expected in the first quarter of 2027. Specifications and cost figures are preliminary.

TAGGED:AI InfrastructureData CentersHardwareLightmatternetworkingOptical InterconnectsPhotonics
SOURCES:Unite AI
Share This Article
Email Copy Link Print
ByAIWadmin
Follow:
Global AI news & information.
Previous Article Adecco spreads a Claude-powered assistant across 40 countries
Next Article Pinecone open-sources a recipe book for vector quantization

You Might Also Like

News

GPT-5.6 Sol deletes user files without warning, users report

By
AIWadmin
News

SpaceX secret foundry targets the part throttling AI power

By
AIWadmin
News

Colossus Promised the Future. Memphis Paid the Price.

By
AIWadmin
News

AI agent builder Yellow.ai lists on Nasdaq through $550M SPAC deal

By
AIWadmin
AIWatcher
Facebook Twitter Youtube Linkedin Rss

Global AI News and Information
AIWatcher is your definitive source for AI updates worldwide, from Silicon Valley to Shanghai.
Our industry coverage keeps you in the loop with the latest news and trends shaping the future of AI.

Quick Links
  • News
  • Articles
  • Spotlight
  • Events
About Us
  • Mission
  • Services
  • Contact
  • Privacy Policy
  • Legal
© 2026 AIWatcher. All Rights Reserved.