Overview
Mithril builds an AI omnicloud platform that aggregates multi-cloud GPUs, CPUs, and storage into a single virtual pool to run distributed AI workloads. Access is through a unified interface (REST API, CLI, and Python SDKs), with options to run jobs at market rates, bid for interruptible spot capacity, or reserve resources at real-time prices. It differentiates itself by spanning multiple providers to create a transparent AI compute market and by offering a managed batch inference service to improve cost-per-task for large AI jobs. The goal is to make AI infrastructure flexible and broadly accessible, turning rigid compute into a fluid, universally usable resource for researchers, startups, and large organizations.
Work here?Claim your company
Simplify's take
What believers are saying
- Mithril launched B200 GPUs on Spot in 2026, expanding premium capacity.
- The June 8, 2026 SkyPilot partnership widens distribution to existing ML workflows.
- Customers already run 10,000+ GPU-hours daily, proving usage depth and retention.
What critics are saying
- AWS, Azure, and GCP can bundle similar GPU marketplaces and crush margins.
- SkyPilot integration from June 8, 2026 weakens switching costs and commoditizes orchestration.
- NVIDIA supply concentration and B200 dependence expose Mithril to sudden capacity shocks.
What makes Mithril unique
- Mithril unifies AWS, Azure, GCP, and bare-metal GPUs behind one pricing layer.
- Its Batch Inference API and Flow SDK target high-throughput AI workloads directly.
- Arc Institute, LG AI Research, and Broad Institute validate real research demand.
Help us improve and share your feedback! Did you find this helpful?
Funding
Industry standards