Rackspace has signed a multiyear agreement with Together AI to build out dedicated Nvidia-powered infrastructure for AI inference workloads, marking the San Antonio company's second major AI infrastructure deal in less than two weeks. The deployment is planned for 2027, but Rackspace says it still needs to secure financing before the buildout can move forward.
A Second Big Bet in Under Two Weeks
According to Rackspace Technology investor relations, the company and Together AI “today announced an initial multiyear strategic AI infrastructure agreement” covering dedicated NVIDIA GB300 NVL72 systems connected with Nvidia Spectrum-X networking to support Together AI's inference workloads. As the San Antonio Business Journal notes, the announcement came “less than two weeks after announcing an AMD-powered partnership with HGS,” making this the second major infrastructure agreement Rackspace has unveiled in rapid succession.
That earlier deal, announced Oct. 29, saw HGS select Rackspace to provide private-cloud and AI-compute infrastructure powered by AMD, with a three-year deployment scheduled to begin in December. Rackspace's own release at the time stated that HGS “has selected Rackspace to provide the private cloud, AI compute infrastructure, powered by AMD,” with the deployment set to begin that month.
Financing Still an Open Question
Deployment of the Together AI infrastructure is planned for 2027, but Rackspace's release is explicit that this timeline is “subject to the satisfaction of certain conditions.” The company said it expects to pursue separate financing to purchase the GB300 NVL72 systems, adding that “no definitive GPU financing commitment has been obtained as of the date of this release.” Rackspace's obligation to actually proceed with the buildout is likewise contingent on a financing condition and other terms spelled out in the definitive agreement.
A Q3 2026 earnings-call transcript obtained by The Motley Fool put a dollar figure on the deal, describing it as having an anticipated total contract value of $380 million “for NVIDIA GB300 infrastructure expected to be in service by mid-2027.” Neither the capacity Rackspace is providing nor the data center where it will be housed has been made public; per Data Center Dynamics, “how much capacity Rackspace is providing to Together AI has not been shared, nor has the data center it will be deployed in.” Separately, Rackspace and Together AI also signed a non-binding letter of intent to explore a joint go-to-market approach for enterprise AI solutions, according to the same investor-relations release.
Part of a Broader AI Infrastructure Push
The Together AI agreement is one of several AI-related moves Rackspace has made in recent months. The same earnings-call transcript described a separate $230 million agreement with Blacksmith for dedicated AMD EPYC bare-metal compute and storage. In September, Rackspace announced it had joined the NVIDIA Cloud Partner Program and unveiled an Institutional Sovereign Pod that, per a GlobeNewswire release from the company, “will bring together the NVIDIA Blackwell architecture with Palantir Foundry and AIP in an environment Rackspace operates.”
Rackspace management has set a target of 20 to 22 megawatts of high-performance AI compute by the end of 2027, projecting annualized run-rate revenues of $420 million to $460 million from that capacity, according to a transcript reported by Benzinga. The push comes as the broader data-center industry scales up at a rapid clip: BloombergNEF estimates capital expenditure among the 14 largest publicly owned data-center operators globally will approach $750 billion in 2026, up from slightly less than $450 billion the year before. The research firm also tracked more than $100 billion in hyperscaler leases for neocloud compute capacity signed over the six months through March 2026, underscoring how competitive the race for AI infrastructure deals has become for companies like Rackspace.
