Rising power and data flows are creating data center thermal friction, a hidden constraint that can undermine the efficiency gains from NVTS and POET technologies.
NVIDIA is investing $4B in transceiver optics with Lumentum and Coherent, strengthening server-to-switch data flow, while GPU-level optical solutions like POET remain an open opportunity.
GPU power delivery feeding compute, illustrating the shift from power chip limits to internal data movement constraints in AI systems