Exposure Candidates

Open-model and agent progress could strengthen Qwen adoption, Alibaba Cloud workloads, and the economics of AI-enabled commerce tools.

Cloud AI revenue growth, model API usage, inference gross margin, and paid enterprise retention.

More reliable multimodal agents could improve Tencent Cloud demand and lower the cost of AI features across games, advertising, and enterprise software.

AI-cloud workload growth, token cost, advertising conversion lift, and enterprise agent retention.

AI cluster density and power demand may support data-center utilization, while serving efficiency could delay some physical capacity additions.

AI customer mix, committed megawatts, move-in timing, utilization, power availability, and return on new capacity.

Larger distributed training and inference clusters can raise demand for higher-speed optical interconnects when network bandwidth becomes the bottleneck.

800G and 1.6T shipment mix, customer concentration, pricing, product yield, and optical attach rate per accelerator.

Falling inference cost can expand AI-cloud usage, but smaller platforms still need utilization and pricing gains to offset capex and supplier dependence.

AI revenue mix, GPU utilization, gross margin, capex intensity, customer concentration, and cash conversion.