Bud Ecosystem launches open-source GPU control plane for cloud services

Bud Ecosystem said Wednesday it is launching Bud GPU Foundry, an open-source control plane that lets operators package mixed GPU fleets as bare metal, VMs, Kubernetes, SLURM and serverless services from one catalog and one bill. The release is aimed at neoclouds and sovereign operators that need to monetize accelerator capacity with tighter billing, isolation and audit controls.

Why it matters: - Bud GPU Foundry is meant to help operators turn mixed accelerator fleets into sellable cloud services without stitching together multiple billing and orchestration systems. - The software targets neocloud operators, sovereign cloud providers and enterprise platform teams that need to serve different customer workloads from the same hardware. - Bud says the platform is designed to make invoices, usage records and isolation rules auditable from one system.

What happened: - Bud Ecosystem announced Bud GPU Foundry as an open-source GPU-as-a-service control plane. - The software is licensed under Apache 2.0 and is installed on an operator’s own hardware. - The platform pools NVIDIA, AMD, Intel, Qualcomm and Huawei accelerators across mixed vendors and generations. - Operators can sell the pooled capacity as bare-metal servers, virtual machines, Kubernetes clusters, SLURM clusters and serverless containers. - Bud GPU Foundry replaces Bud Pod as the second of the nine layers of the Bud Novaria AI OS. - Bud Ecosystem said the platform is in production at a sovereign cloud operator in India.

The details: - A customer can request a specific configuration, such as four accelerators with 80 GB of memory each and a high-bandwidth interconnect. - The platform assigns capacity using a published rule and places workloads by network topology on one node or across many. - Isolation levels range from a whole bare-metal host to a hardware partition of a single accelerator where the hardware supports it. - Customers that may compete with each other receive only isolation grades enforced in hardware or in the network fabric. - Metering begins only when the capacity is healthy and responding. - Failed startup attempts are recorded at zero cost with a failure reason. - The usage screen and the invoice come from the same record. - The platform supports prepaid billing, contract billing, spend caps, reseller organizations and white-label branding. - An enterprise platform team can use the same catalog and metering system to allocate capacity internally and charge it back to business units. - Air-gapped deployment is a standard configuration, and no component needs the public internet.

Between the lines: - Bud is positioning GPU infrastructure as a governed utility, not just a pool of expensive hardware. - The open-source angle is meant to address trust concerns in billing and capacity allocation, especially where operators do not want proprietary black-box systems deciding customer charges. - Bud AI Foundry runs on Bud GPU Foundry as an ordinary tenant, so token-based AI services inherit the isolation, quotas and audit trail of the underlying capacity layer. - Capacity and token usage are metered separately but roll up through one account structure, which Bud says lets operators measure margin on each endpoint. - Model training, data grounding, guardrails and agents all run on the same fleet under the same permissions. - Kevin Johnson, co-founder and COO of Bud Ecosystem, said neocloud operators are stuck choosing between customer demand and operational reality, and said Bud GPU Foundry bridges that gap with an open orchestration layer that turns idle capacity into billable compute.

What's next: - Bud GPU Foundry is available now. - Bud said the source code will be published under Apache 2.0, and the repository link will be posted on the product page. - Operators can buy professional services and support if they run the software on their own. - The software is also included in the Bud Novaria AI OS license. - Bud points to a proof of concept that registers one cluster, onboards two customers with different isolation requirements and compares the first invoice with the usage record. - More information is available at the product page.

The bottom line: - Bud is trying to make GPU fleets easier to sell, meter and govern by wrapping them in one open-source control plane instead of multiple fragmented stacks.