VESSL Cloud GPU Cloud for ML Teams

VESSL AI, Inc. is a San Francisco-registered company with a Korean entity in Seoul that started as an MLOps platform for AI developers and raised a 12 million USD Series A in October 2024. Its goal is to give machine learning teams fast access to GPUs without managing Kubernetes or cloud setups themselves.

In February 2026 the product was relaunched as VESSL Cloud, a fully managed GPU cloud. Teams open workspaces with JupyterLab or SSH, run one-command batch jobs, pause and resume while only storage is billed, and switch GPU types without rebuilding environments. VM clusters offer InfiniBand networking and dedicated storage, and Provisioned Throughput serves models through OpenAI- and Anthropic-compatible APIs. Capacity spans Korea, the US and Europe. Users include Hyundai Motor, KT, Hanwha Life, Upstage and research groups at Stanford and KAIST.

• Container Compute: Jupyter, SSH and IDE workspaces plus one-command batch jobs.
• VM Cluster: multi-GPU clusters with 400 Gbps InfiniBand and up to 200 TiB dedicated NFS.
• Provisioned Throughput: model serving through OpenAI- and Anthropic-compatible APIs with contractual targets.
• Shared storage: team volumes with organization- and team-level access control.
• Multi-region routing: capacity across Korea, the US and Europe with automatic fallback.
• GPU lineup: A100, H100 and L40S on demand; H200 and Blackwell on reservation.

Vietnamese AI startups can fine-tune Vietnamese language models on H100s billed by the second. University labs can give students JupyterLab workspaces that pause when idle. Automotive and robotics teams can run large training jobs on reserved multi-GPU clusters.

Per-second billing and pause-and-resume workspaces mean teams pay only for compute they use. Listed on-demand prices start at 1.48 USD per hour for A100 and 2.98 USD per hour for H100. No cloud or Kubernetes setup is needed, so researchers start in minutes. Hyundai Motor's autonomous driving team reports cutting deployment time from five months to one week.

Demand for GPU capacity for training and fine-tuning models is outpacing supply, and smaller AI teams want on-demand access without long hyperscaler contracts. Specialized GPU clouds with simple tooling are growing as Asian companies and universities train local-language and physical AI models.

Legal entities: VESSL AI, Inc. in San Francisco and VESSL AI KOREA Inc. in Seoul. Raised a 12 million USD Series A in October 2024 led by A Ventures, bringing total funding to 16.8 million USD. States SOC 2 Type II certification. The original VESSL MLOps platform remains available to existing users.