What changed
Rackspace and Together AI signed an initial multiyear agreement for Rackspace to operate dedicated NVIDIA GB300 NVL72 systems with Spectrum-X networking for Together AI inference. Deployment is planned for 2027 and includes facility, liquid cooling, networking, storage and managed-cloud operations.
Why it matters
The agreement is another concrete move toward dedicated inference capacity operated outside a hyperscaler, with data-sovereignty and regulated-enterprise positioning. Its value depends on delivered power, equipment, financing and customer-access terms rather than the announcement itself.
What to do
Treat the capacity as future supply. Ask for financing closure, facility and power milestones, system count, acceptance tests, supported models, service terms, residency, price and an independent production benchmark before allocating workloads.
Caveat
The deployment is not live. Rackspace says its obligations depend on financing, facilities, power and equipment; GPU financing was not finalized. A separate joint go-to-market letter is non-binding and future controls or configurations require later agreements.