Infrastructurenvidiacorningoptical fiberco packaged optics

Nvidia partners with Corning to expand optical manufacturing

||By LDS Team
7.1
Relevance Score
Nvidia partners with Corning to expand optical manufacturing
Photo: CNBC · rights & takedowns

CNBC reports that Corning and Nvidia announced a multiyear partnership to build three new U.S. advanced manufacturing plants dedicated to optical technologies for Nvidia, the companies said in a joint press release reported by CNBC. CNBC reports the facilities are expected to create at least 3,000 jobs and to increase Corning's U.S. optical manufacturing capacity by 10-fold. Financial terms were not disclosed, CNBC adds. Corning's corporate blog (Sean Kelly, March 18, 2025) frames the work as part of a broader move toward co-packaged optics, and describes Corning solutions designed for high-density, low-power fiber interconnects compatible with next-generation switch architectures.

What happened

CNBC reports that Corning and Nvidia announced a multiyear collaboration to build three advanced manufacturing plants in North Carolina and Texas dedicated to optical technologies for Nvidia, according to a joint press release quoted by CNBC. CNBC reports the expansion is expected to create at least 3,000 jobs and to increase Corning's U.S. optical manufacturing capacity by 10-fold. CNBC reports that financial terms were not disclosed. CNBC also quotes Corning CEO Wendell Weeks: "What Nvidia is doing is nothing short of extraordinary, not just for the future of AI, but for the American advanced manufacturing workforce."

Technical details

Editorial analysis - technical context: Public reporting and Corning documentation position co-packaged optics as the core technical theme. Per Corning's blog post by Sean Kelly (March 18, 2025), Corning describes high-density rack optical cabling, low outer-diameter high-fiber-count cables, factory-terminated MMC connectors, and solutions intended to bring fiber connectivity closer to photonic chips. Industry literature frames co-packaged optics as a way to reduce power and thermal load from electrical copper links by moving optical interfaces closer to switching ASICs, improving bandwidth per rack and lowering energy per bit.

Context and significance

The tie-up combines a leading GPU and systems vendor with a legacy optical materials and cabling manufacturer at a moment when hyperscaler and AI cluster networking needs are rapidly growing. Public coverage links the deal to broader efforts to replace high-density copper interconnects with fiber in GPU racks, which industry analysts have argued is necessary to scale throughput while managing power and cooling. Corning's recent stock gains, noted by CNBC, reflect its expanded role supplying optical components to cloud and AI infrastructure customers.

What to watch

For practitioners: observers should track:

  • technical specifications and form factors for the co-packaged optics components that emerge from the collaboration
  • interoperability standards and connector choices that affect deployment and field service
  • supply-chain timelines and factory output schedules in Corning announcements versus customer procurement windows. Monitoring how major OEMs and hyperscalers reference optical versus electrical topologies in upcoming system briefs will indicate the speed of adoption

Key Points

  • 1Large-scale U.S. optical capacity expansion supports higher-density AI racks, lowering energy per bit and easing thermal constraints.
  • 2Corning-Nvidia collaboration centers on co-packaged optics, a technical path often cited to replace copper in high-throughput GPU interconnects.
  • 3A 10-fold capacity increase and 3,000 jobs signal industrial-scale supply-chain investment, which matters for procurement lead times and deployment planning.

Scoring Rationale

The deal materially affects AI infrastructure supply capacity and the technical trajectory toward `co-packaged optics`, a near-term enabler for denser GPU clustering. It is important for practitioners planning deployments but not a paradigm-shifting research release.

Practice interview problems based on real data

1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.

Try 250 free problems