Rackspace Surges 20% Premarket After Securing 30MW AI Computing Contract From AMD
Miles Bennett
Rackspace signed a 30-megawatt AI compute deployment deal with AMD spanning late 2026 through 2028, sending its stock up roughly 20% pre-market — the legacy cloud host is using one contract and one round of layoffs to rebrand itself as an enterprise AI infrastructure player.
What does this contract actually buy?
Rackspace will deploy 30 MW of dedicated AMD AI compute across its global data centers in phases, covering AMD Instinct GPUs (MI355X, MI350P, and successors) plus EPYC CPUs.
In plain terms = 30 megawatts is roughly a mid-sized data center's worth of power, all earmarked for AI workloads — this is production-scale capacity, not a pilot.
Deployment starts in late 2026 and runs through 2028, converting an earlier memorandum of understanding into a binding agreement.
Why target regulated industries?
Both companies are positioning the capacity for regulated enterprise workloads; healthcare clients have already shown early interest in clinical AI and large-scale inference.
This means → Rackspace is not trying to steal general cloud share from AWS or Azure. It is betting on a niche: enterprises too compliance-heavy to drop sensitive workloads onto a public cloud.
AMD SVP Dan McNamara said the goal is to help these enterprises deploy AI infrastructure in an "open, scalable, and accountable" manner.
Cutting 15% of staff — how steep is the pivot?
Alongside the contract, Rackspace announced a workforce restructuring that will cut roughly 15% of its global headcount. Most affected employees were notified around June 10; the rest will exit over the following six months.
The cuts target legacy service-delivery roles, concentrated in the public-cloud business unit — the company is explicitly reducing its dependence on that segment.
This means → Rackspace's calculation is blunt: the headcount serving traditional hosting is no longer worth keeping. The savings go straight into AI.
Do the savings cover the bet?
One-time restructuring charges are estimated at $14 million to $19 million, mostly recognized within 2026.
Once fully implemented, the company expects annual run-rate savings of $75 million to $85 million.
In plain terms = spend under $20 million to save roughly $80 million a year — a payback period of under three months on paper. A "significant portion" of the savings will be reinvested into AI engineering and infrastructure.
How did the market vote?
After the announcement, Rackspace jumped roughly 20% pre-market; AMD rose about 1%.
This reflects a repricing narrative for Rackspace: from a shrinking legacy cloud host to a turnaround story with a real AI order book.
AMD's 1% move tells you that 30 MW is incremental but not needle-moving at AMD's scale — the company truly being repriced here is Rackspace itself.
Content is for reference only, not financial advice.