A Chinese startup, co-founded by the former Vice President of Technology at Alibaba Cloud, has caught the attention of NVIDIA, which reportedly acquired a stake in the company for several hundred million dollars. The company’s name is Lepton AI, and it offers a platform designed to facilitate artificial intelligence workloads across diverse GPU resources.
Rumors about this acquisition first surfaced in late March and early April. While there has been no official confirmation from either party, the startup’s website now redirects visitors to NVIDIA’s official catalog. More specifically, it points to a product available in early access called DGX Cloud Lepton.
NVIDIA Aims to Bring Together Major Cloud Infrastructure Players
Workspaces, private registries, observability tools — NVIDIA’s platform appears to follow the paradigms established by AI development environments like SageMaker or Vertex AI. However, NVIDIA emphasizes not just the software side of things, but also its marketplace for GPU resources. The company has announced around fifteen partners so far, including CoreWeave, Fluidstack, Foxconn, Lambda, and SoftBank. They have promised additional leading cloud providers will join in the future.
This ecosystem is integrated within NVIDIA’s broader stack of tools and frameworks, including NIM, NeMo, and Cloud Functions. It offers flexible deployment options, allowing users to opt for on-demand usage or reserved instances. Users can select their preferred geographic regions and connect their own hardware resources, subject to certain conditions — for example, they must use at least Ubuntu 22.04 LTS. The management of the system is handled by GPUd, a system used notably by companies like Meta and Uber.
