API
< Back to situations

Monitor this situation.

[SITUATION] · [QUIET] · [TECHNOLOGY]

16 clusters · 51 sources · 87 days · First seen · Last updated

AI infrastructure connectivity and optical technology trends

Overview

The focus of AI infrastructure development is shifting from the procurement of GPUs and High Bandwidth Memory (HBM) toward the connectivity between components. As AI data centers expand, traditional copper-based electrical connections are facing limitations regarding power efficiency and signal attenuation, a challenge often referred to as the “copper wall.” To address this, optical communication technology is emerging as a critical solution, involving a transition from long-distance rack-to-rack connections to Co-Packaged Optics (CPO) and the eventual optical conversion of GPU chip interfaces.

Recent industry developments highlight advancements in both hardware architecture and manufacturing testing. Nvidia has unveiled a prototype of its Optics-on-Interposer (OOI) technology, a CPO design that moves optical communication circuitry directly onto the silicon interposer to increase bandwidth density and reduce power consumption. Complementing this, Voyant Photonics is developing a chip-scale optical circuit switching (OCS) engine to reduce optical-electrical-optical conversions. To support commercialization, new testing technologies are emerging, such as Pantech’s MSS HG for silicon photonics analysis and NightJar technology, which uses hyperspectral imaging for rapid wafer-level screening.

Newer innovations focus on reducing component footprint and energy use. Researchers at Waseda University have demonstrated a 180 Gbps optical transmitter that eliminates the need for traditional optical isolators by coordinating the responses of a tunable laser, silicon nitride ring resonators, and a thin-film lithium niobate modulator. Additionally, Ciena is developing Vesta-Engines, which use silicon photonics to place optical components closer to computing chips like ASICs. This linear architecture operates without local digital signal processors or retimers, aiming to increase connection density while allowing more power to be dedicated to computing units.

Entities

Nvidia · Cisco · TrendForce · Goldman Sachs · OpenLight

Timeline

  1. [TECHNOLOGY] 3 sources
    Optical technology advances aim to optimize AI data center connectivity

    New optical transmitter designs and the rise of VCSEL Near-Packaged Optics (NPO) are set to address the growing connectivity and power demands of high-speed AI data centers.

  2. [TECHNOLOGY] 2 sources
    Silicon Photonics: New Testing Technologies Aim to Accelerate Commercialization

    New silicon photonics testing tools like Pantech’s MSS HG and NightJar technology aim to boost yields and speed up commercialization for AI and high-speed optical communication markets.

  3. [TECHNOLOGY] 2 sources
    Silicon photonics and BTO materials target datacenter optical interconnects

    Advancements in silicon photonics and barium titanate integration aim to meet the high bandwidth and low latency demands of AI and cloud computing by improving datacenter optical interconnects.

  4. [TECHNOLOGY] 4 sources
    AI Infrastructure Drives Growth in PCB and Photonic Packaging Markets

    AI and data center expansion is driving massive growth in AI server PCB and photonic packaging markets, with significant technological shifts toward optical interconnects and high-layer boards.

  5. [TECHNOLOGY] 2 sources
    Nvidia and Voyant Photonics introduce optical technologies for AI data centers

    Nvidia has prototyped Optics-on-Interposer technology to boost AI bandwidth, while Voyant Photonics is developing a chip-scale optical circuit switching engine for data centers.

  6. [TECHNOLOGY] 2 sources
    Researchers achieve breakthroughs in optical interconnects for AI computing

    New breakthroughs in GaN microLED bandwidth and silicon optical modulators aim to enhance data transmission speeds and energy efficiency for AI data centers.

  7. [TECHNOLOGY] 8 sources
    NVIDIA and Cisco announce USFP optical modules for AI infrastructure

    NVIDIA and Cisco announced USFP, a new high-density optical module for AI infrastructure providing 400Gb/s per lane and up to 12.8Tb/s capacity with improved power efficiency.

  8. [TECHNOLOGY] 3 sources
    Advancements in silicon photonics drive high-speed, low-power optical communication

    Waseda University has developed a low-power 180 Gbps optical transmitter for AI data centers, while OKI and Fraunhofer HHI have successfully prototyped a hybrid integrated laser vibrometer module.

  9. [TECHNOLOGY] 2 sources
    AI infrastructure demand drives shift to 400G and 800G networking

    AI growth is driving massive demand for high-speed 400G and 800G network infrastructure, prompting major investments in optical fiber and specialized power solutions for data centers.

  10. [TECHNOLOGY] 10 sources
    Optical fiber demand surges due to AI data center expansion

    AI and data center growth is driving massive demand for optical fiber, causing Sterlite Technologies' stock to surge and fueling advancements in low-latency, long-distance optical communication.

  11. [TECHNOLOGY] 4 sources
    Optical interconnect technologies advance to support AI scaling

    Companies including LightSpeed Photonics and ams OSRAM are advancing optical interconnect technologies, such as Thin-Film VCSELs and 400-Gbps components, to solve data bottlenecks in AI infrastructure.

  12. [TECHNOLOGY] 2 sources
    AI infrastructure growth drives demand for advanced optical interconnects

    AI infrastructure growth is driving a shift toward advanced optical interconnects and co-packaged optics to manage the high speed, scale, and power demands of massive GPU clusters in data centers.

  13. [TECHNOLOGY] 2 sources
    AI infrastructure demand drives growth in optical networking and routing

    AI infrastructure expansion is driving a surge in optical networking and high-end routing revenue, with hyperscalers and cloud providers fueling significant growth for companies like ADTRAN.

  14. [TECHNOLOGY] 3 sources
    Optical networks become critical infrastructure for AI

    The AI revolution is driving a shift from copper to optical networks in data centers to overcome energy and speed bottlenecks, with high-speed optical transceiver use projected to exceed 60 percent by 2026.

  15. [TECHNOLOGY] 2 sources
    AI demand drives optical component shortages and photonic innovation

    AI-driven demand is causing acute optical component shortages, forcing data centers to diversify suppliers and adopt new photonic technologies to overcome electrical and bandwidth bottlenecks.

  16. [TECHNOLOGY] 3 sources
    AI infrastructure faces power and fiber-optic bottlenecks

    AI expansion is hitting critical infrastructure limits, specifically regarding power availability and the need for advanced fiber-optic connectivity to replace traditional copper wiring.

Sources

advisoranalyst.com · ascii.jp · biz.heraldcorp.com · blogs.itmedia.co.jp · compoundsemiconductor.net · cryptobriefing.com · digitaljournal.com · dreamnews.jp · eetasia.com · eetimes.jp · electronicsb2b.com · electronicsweekly.com · finance.technews.tw · finanzen.at · finanzen.net · folgefonnsenteret.no · futurumresearch.com · gcn.com · integratormedia.com · interestingengineering.com · investorplace.com · ithome.com · knowridge.com · ledinside.com · leiphone.com · lightwaveonline.com · ltvnews.net · merfoldko.hu · mimemo.io · newspaceeconomy.ca · nextbigfuture.com · nextplatform.com · photosbysuzanne.com · prtimes.jp · quantumzeitgeist.com · rocketnews.com · sharedeals.de · siliconsemiconductor.net · student-circuit.com · techblog.comsoc.org · the-miyanichi.co.jp · thepatent.news · theregister.co.uk · theregreview.org · tokenpost.com · tw.nextapple.com · wirecable.in · world-today-news.com

This summary has been updated 17 times: see revision history