Global flagships × full-stack domestic chips, covering everything from trillion-parameter training to high-throughput inference. Pay as you go, delivered in minutes.
An integrated base of network, storage, orchestration and compliance.
Mixed pool of B200/GB200 and domestic chips, scale on demand.
In-region deployment to meet data and compliance needs.
Low-latency network across NA, EU, Middle East and SEA.
40-60% lower TCO thanks to China's compute scale.
NVIDIA's latest architecture alongside self-reliant domestic silicon.
NVIDIA Blackwell B200, GB200 NVL72, plus Ascend 910C, Cambricon MLU590, Hygon DCU and an H800/5090/4090 heterogeneous pool.
Prioritize domestic chips for sovereignty and compliance; choose NVIDIA flagships for peak performance and ecosystem maturity.
Both pay-as-you-go (Serverless per token/second) and monthly/annual reservation.
Yes, in-region and designated-region deployments are supported.