Nvidia Commits $6.5B to Photonics Over Copper
Key insights
- Nvidia deployed at least $6.5 billion across five photonics deals since March 2026, spanning the full optical interconnect supply chain.
- Copper interconnects face prohibitive power and throughput limits at million-GPU AI cluster scales, making optical transmission the likely successor technology.
- Nvidia's Ayar Labs participation signals interest in silicon photonics integrated directly on chip packages, extending beyond external optical cabling.
Why this matters
Optical interconnects are now a confirmed strategic priority for the world's dominant AI chip vendor, which means every hyperscaler and data center operator will face pressure to plan for photonics-based fabric within the next hardware refresh cycle. Nvidia's vertical integration play mirrors the CUDA strategy: if it also controls the physical transmission layer, customer switching costs increase substantially beyond the GPU itself. Companies building AI infrastructure today need to evaluate whether their networking vendors have credible photonics roadmaps, since copper-optimized architectures may require expensive retrofits as cluster sizes scale past current thresholds.
Summary
Nvidia has assembled a $6.5 billion photonics portfolio since March 2026, targeting the bottleneck that stops copper-wired AI clusters from scaling.
As clusters approach million-GPU configurations, copper hits hard limits on power and throughput. Optical moves data via light at higher bandwidth and lower energy draw, making it the only viable path at that scale.
Essentially: (Lumentum, Coherent, Marvell, Corning, Ayar Labs) are the vendors Nvidia is backing as the next physical layer of AI compute.
- $2B deployed across Lumentum, Coherent, and Marvell; $500M Corning deal; participation in Ayar Labs' $500M Series E.
- Analysts frame this as Nvidia locking in the interconnect stack before copper constraints force customers to scramble for alternatives.
If optical becomes the hyperscale default, Nvidia's role as both chip supplier and infrastructure architect compounds well past its current GPU footprint.
Potential risks and opportunities
Risks
- If photonics manufacturing yield rates do not improve materially by 2028, Nvidia's investment thesis stalls and hyperscalers face a gap with no viable copper replacement at million-GPU scale.
- Coherent and Lumentum, both publicly traded, could face margin pressure if Nvidia's investments come with pricing concessions, negatively affecting returns for existing shareholders in the near term.
- Competing optical interconnect standards backed by AMD or Intel could fragment the ecosystem, stranding Nvidia's preferred vendors and making portions of the $6.5 billion commitment illiquid.
Opportunities
- Optical networking equipment vendors Ciena and Infinera gain leverage in data center refresh conversations as Nvidia's investments validate the photonics roadmap at the highest level.
- Data center operators that qualify photonics suppliers now, before supply constraints emerge, can negotiate preferred pricing ahead of a likely 2027-2028 demand spike.
- Silicon photonics startups including Rockley Photonics and HyperLight become credible acquisition targets as hyperscalers seek to replicate Nvidia's supply-chain positioning in optical interconnects.
What we don't know yet
- Breakdown of equity versus supply-agreement terms in the $2 billion Lumentum/Coherent/Marvell allocation has not been disclosed publicly.
- Whether Nvidia's investments include exclusivity or supply-priority clauses that would disadvantage AMD and other chip vendors at scale is unconfirmed.
- Timeline for commercial availability of Ayar Labs' TeraPHY silicon photonics chiplets in production Nvidia systems remains unspecified beyond a general 2027 projection.
Originally reported by cnbc.com
Read the original article →Original headline: Nvidia Has Committed $6.5 Billion to Photonics Companies Since March — Bet That Optical Interconnects Will Replace Copper as AI Data Center Scale Hits Physical Limits