Hundreds of organizations including Nubank, Abridge, Applied Compute, and H Company use SkyPilot to turn fragmented clouds, clusters, and accelerators into one unified AI supercomputer
SkyPilot announced it has launched from stealth with $20 million in seed funding. The company introduces SkyPilot Platform, a unified AI compute platform that helps frontier AI teams manage their AI compute across hyperscalers, neoclouds, Kubernetes clusters, and accelerator types.
The funding round was led by Lux Capital, with participation from Amplify Partners, Coatue Management, Foundation Capital, Race Capital, The House Fund, and leading technology operators including Ali Ghodsi, CEO of Databricks; Jeff Dean, Chief Scientist at Google; Guillermo Rauch, CEO of Vercel; Amjad Masad, CEO of Replit; Clem Delangue, CEO of Hugging Face; and Tristan Handy, CEO of dbt Labs.
SkyPilot was founded by Berkeley researchers Zongheng Yang, Zhanghao Wu, and Romil Bhardwaj alongside Databricks co-founder Ion Stoica and networking and cloud computing pioneer Scott Shenker, both Berkeley professors. The technology behind SkyPilot emerged from years of AI systems research at UC Berkeley.
Read More:Â SalesTechStar Interview with Haley Gault, Senior Account Executive, Salesforce
Custom Intelligence Is Existential but Stalled by Fragmented AI Compute
From frontier AI labs to Fortune 500 enterprises, organizations are increasingly building custom intelligence: agents, applications, and post-trained models built on their own data and running on infrastructure they control.
Building custom intelligence at scale requires more compute than any single provider can offer. AI teams are assembling infrastructure across hyperscalers, neoclouds, Kubernetes clusters, and multiple generations of accelerators. As those environments become more distributed, managing AI compute has become a complex operational challenge that slows the development of custom intelligence.
SkyPilot turns fragmented clouds, clusters, and accelerators into one unified AI supercomputer. Through a unified control plane, organizations can develop, manage, monitor, and scale AI workloads across providers without changing how those workloads are built or operated. Today, SkyPilot powers AI compute across hundreds of organizations. Its open source project has surpassed 14 million downloads and attracted more than 280 contributors.
“Every organization is building custom intelligence around its own data and domains,” said Zongheng Yang, CEO and co-founder of SkyPilot. “The challenge is that the AI compute needed to build it is fragmented across clouds. SkyPilot gives frontier AI teams a single platform to manage that infrastructure so they can build custom intelligence faster.”
Read More:Â Break Your Bias: The Cost Of Cutting Corners When Choosing A CRM
Introducing SkyPilot Platform
SkyPilot Platform is a managed AI compute platform for organizations operating large-scale AI infrastructure. Built on top of the SkyPilot open source project, it provides large GPU fleet operations, frontier workload support, standardized cluster management, workload orchestration, governance, and enterprise controls through a single platform.
Once customers bring their AI compute, the platform supports development, agentic workloads, training, reinforcement learning, inference, evaluations, and multi-cluster production serving. Customers also gain access to fleet-wide GPU health monitoring, automated remediation, high availability, team and quota management, single sign-on, and SOC 2 compliance.
Private preview customers have successfully used the platform to manage over 10,000 GPUs and support over 200 researchers per organization. Some saw performance improvements of up to 20x compared to SkyPilot open source thanks to the platform’s optimizations.
