As AI model training matures, inference computing power is becoming the central focus. Low-latency, edge-side, and high-density infrastructure are becoming key construction priorities.
Driven by the exponential growth of AI training clusters scaling to hundreds of thousands of GPUs, the landscape of transmission distance, bandwidth, and power consumption within data centers has been fundamentally redefined.
Nvidia injected $2 billion each into photonics manufacturers Lumentum and Coherent, securing future production capacity and technology rights for advanced optical components with a total investment of $4 billion.