Skip to content
AI.info

The Pulse

Rackspace and Together AI plan dedicated GB300 inference deployment

Rackspace and Together AI signed a multiyear agreement for dedicated NVIDIA infrastructure, with deployment planned for 2027. The project remains subject to financing and other conditions, while a separate non-binding letter explores further cooperation.

Rackspace and Together AI plan dedicated GB300 inference deployment

Rackspace and Together AI have signed a multiyear agreement for dedicated NVIDIA infrastructure to support Together AI inference workloads, Rackspace announced on November 10. Deployment is planned for 2027, subject to financing and other conditions.

The agreement is binding, but Rackspace’s obligations also depend on the availability of facilities, power and equipment on the expected timeline. On an earnings call, Rackspace Chief Executive Officer Gajen Kandiah put the anticipated total contract value at $380 million over several years, subject to conditions, according to a transcript published by Stock Analysis. He said the infrastructure was expected to be in service by mid-2027.

Infrastructure and platform roles

The planned deployment uses NVIDIA GB300 NVL72 systems connected through NVIDIA Spectrum-X Ethernet networking. Rackspace expects to supply and operate the infrastructure, including power, liquid cooling, networking and storage. The systems are planned for a purpose-built, liquid-cooled facility.

Together AI expects to provide the managed cloud platform running on that infrastructure, with access to open-source models, inference optimizations and developer tools. The parties also expect to evaluate additional security, governance and compliance controls for future offerings, depending on customer requirements and agreed configurations.

Rackspace intends to seek contractual commitments for dedicated capacity and manage related financing and capital commitments with the goal of aligning them to expected customer demand and the useful life of the hardware. The approach is part of its plan to support AI platform companies, systems integrators and enterprises as they scale infrastructure.

Financing remains open

Rackspace expects to seek separate financing to purchase the GB300 NVL72 systems and is discussing it with potential lenders, but had no definitive GPU financing commitment as of the announcement. The company cautioned that financing might not be obtained on acceptable terms, or at all.

A separate share purchase agreement effective November 6 provides for an aggregate $66 million private placement of Rackspace common stock with an investor group led by Voss Capital. The company expects to use the net proceeds for general corporate purposes, working capital, capital expenditures and growth capital, including funding for GPU-related initiatives.

Further cooperation under discussion

The companies also signed a separate, non-binding letter of intent to explore a joint approach to enterprise-focused AI offerings. That letter contemplates evaluating how Together AI’s inference, post-training and reinforcement learning capabilities could be combined with Rackspace infrastructure, platforms and engineering services. Any joint offerings would require further agreement and would depend on customer requirements.

Kandiah said the letter also covers potential future capacity expansion. It is not a definitive agreement and remains subject to future agreements and other conditions.

The companies also intend to explore opportunities to serve enterprise customers seeking greater control, compliance and data sovereignty. Possible configurations could bring Together AI’s cloud closer to enterprise data and workloads. Their availability and features, including data residency, security and compliance, would depend on the specific offering and customer arrangement.

Sources

Related stories

Explore

More articles