Rackspace is slashing 15% of its workforce and partnering with AMD to deploy a 30-megawatt artificial intelligence computing cluster, the company announced Tuesday. The dual moves signal a sharp strategic shift as the cloud services provider bets heavily on AI infrastructure to revive its business.
A costly pivot
The layoffs, affecting roughly 15% of employees, are meant to free up capital for the AI investment. Rackspace didn't specify exact headcount numbers, but the company employed about 8,600 people at the end of last year. That would put the cuts at roughly 1,300 roles.
This isn't the first restructuring for the San Antonio-based firm. Rackspace has been trimming headcount on and off since its 2016 go-private deal. But the current round is explicitly tied to a single bet: build out big AI compute capacity and hope clients flock to it.
The company framed the reductions as necessary to "fund the pivot to AI infrastructure." Workers in non-AI business units are expected to bear the brunt.
The AMD partnership
On the technology side, Rackspace is going with AMD rather than the more common Nvidia GPUs for the 30MW cluster. The chips — AMD's Instinct MI300X accelerators — will power what Rackspace calls a "private cloud AI supercomputer." The 30MW figure refers to the total power draw of the system, a common metric for large-scale compute deployments.
AMD has been making inroads into data center AI, though it still trails Nvidia in market share. For Rackspace, using AMD could mean lower costs or better availability than Nvidia's supply-constrained hardware. The company didn't say when the cluster will go live or which customers have signed on.
What gets left behind
The layoffs raise questions about Rackspace's legacy hosting and managed cloud businesses. The company built its name on managing other people's servers — first physical boxes, then AWS and Azure environments. That traditional business has faced pricing pressure from bigger rivals and automation from the hyperscalers themselves.
Now Rackspace is essentially saying the future is in building and running custom AI hardware for clients, not in general-purpose cloud management. The 15% reduction suggests the shift will be painful for those in older service lines.
Rackspace didn't detail severance packages or which locations will be hit hardest. The company has offices in the U.S., U.K., Australia, and India.
What remains unclear is whether the AI compute bet will pay off fast enough. The 30MW cluster is a sizable deployment, but it's tiny compared to what Microsoft, Amazon, or Google are building. Rackspace will need to convince enterprises it can offer more flexibility or better support than the big three — all while running a leaner operation.
