Nvidia Drops $4 Billion on Silicon Photonics as It Locks Down Next-Gen Chip Interconnects
Coherent and Lumentum each receive $2 billion in cash plus multi-billion-dollar purchase commitments
The deals, announced Monday, represent Nvidia's latest move to vertically integrate its AI infrastructure supply chain. Each company receives $2 billion in direct investment along with multi-billion-dollar purchase commitments, giving Nvidia priority access to the photonics components that enable high-bandwidth data transfer between GPUs in massive AI training clusters.
Silicon photonics uses light rather than electrical signals to move data between chips, offering dramatically higher bandwidth and lower power consumption. As AI models grow larger and training clusters scale to tens of thousands of GPUs, the interconnects between those chips become a critical bottleneck.
The investment also furthers Nvidia's push to build out American manufacturing capacity for these components, aligning with broader industry efforts to reduce dependence on overseas supply chains for critical semiconductor technologies.
Analysis
Why This Matters
Nvidia isn't just buying chips anymore — it's buying the plumbing. Silicon photonics is the unsexy infrastructure layer that determines how fast AI clusters can actually operate, and Nvidia is making sure no competitor can outbid it for supply.
Background
As AI training clusters scale past 100,000 GPUs, traditional copper interconnects hit bandwidth and power walls. Photonics solves this with light-based data transfer, but manufacturing capacity is limited. Nvidia's $4 billion bet ensures it gets first dibs.
What to Watch
Whether AMD and other competitors can secure their own photonics supply, and whether this vertical integration trend accelerates across the AI chip industry.