Rackspace's AI Pivot: A $75M Bet on Sovereign Inference

Rackspace doubles down on controlled, sovereign AI with AMD and Palantir, betting on a $15-20M per MW revenue model and 50%+ EBITDA margins.

RXT · Earnings Call · 2026-08-11

The Strategic Pivot Becomes Concrete

Rackspace has been telegraphing a pivot toward enterprise AI for over a year, but the second quarter turned ambition into a definitive plan. The company signed a binding agreement with AMD to deploy 30 megawatts of AMD-based compute across its data centers, starting with a ~2MW footprint that should be live by the end of 2026. The economics are hard to ignore: management expects $15-20 million in revenue per megawatt, translating to $450-600 million of annual revenue at full build-out, with EBITDA margins north of 50%. To fund this, Rackspace plans to use OEM and asset-backed financing, preserving its cash for operations. As CEO Gajan Kandiah put it, “We are evaluating financing for a significant portion of the compute hardware through a combination of OEM financing, equipment financing and other asset-backed credit facilities with the financing collateralized by the newly acquired hardware.” — Gajen Kandiah, Chief Executive Officer · 2026-08-11 This is not a typical capex-heavy neocloud buildout; it's a measured, capital-light entry into the AI infrastructure game, leveraging existing data centers and power capacity.

The attractive economics of our new growth vector bears repeating. Capital expenditures for the first deployment are expected to be approximately $75 million... We expect to average $15 million to $20 million in revenue per megawatt deployed... This range translates to $450 million to $600 million in annual revenue for the full 30-megawatt deployment... EBITDA margins in enterprise AI to be in the 50% plus range.
Gajen Kandiah, Chief Executive Officer · 2026-08-11

The AMD deal is tightly coupled to the company's long-standing relationship with Palantir. Management described a boot-camp-to-agent-camp motion that's now converting into production deals across regulated industries. "The Palantir partnership over the last 4 to 5 months has truly gained traction," Kandiah said on the call, “especially as I think customers have started or enterprise customers have started to understand the importance of what I would call sort of institutional sovereignty in terms of where models run.” — Gajen Kandiah, Chief Executive Officer · 2026-08-11 The same theme of sovereignty—extending beyond national borders to the enterprise—was echoed by Kandiah's follow-up: “I wouldn't call it evolution anymore. I think this is an emerging trend around sovereign AI and the concept of sovereignty going beyond sort of country boundaries down to enterprise boundaries.” — Gajen Kandiah, Chief Executive Officer · 2026-08-11 This is not just a marketing narrative. Rackspace is repositioning itself as the accountable operator of the full enterprise AI stack—core, cloud, and edge—and its strategic pivot is now backed by concrete milestones and a hired leader for AI infrastructure. The company also released RackAI, its inference and fine-tuning platform, and is expanding its funded engagement offerings across AWS and Microsoft to convert enterprise AI ambition into governed production deployments. The pivot is squarely aimed at inference workloads—a category that's expected to dominate AI compute by 2030.

Financial Discipline Amid Transformation

The pivot comes with a deliberate cost: revenue guidance for 2026 calls for a 7% decline at the midpoint, driven by the intentional exit of low-margin Public Cloud resale. Second-quarter revenue of $670 million was up 1% year-over-year, but the beat was largely a timing artifact in Private Cloud. Gross margin fell 120 basis points to 18.6%, and non-GAAP operating profit was flat at $27 million. Management reiterated full-year adjusted EBITDA of $285-295 million and expects free cash flow to turn positive in the second half. The company's balance sheet remains stretched—liabilities-to-assets stood at over 140%—but CFO Mark Marino emphasized a deleveraging mindset, having repurchased $11.2 million of senior notes during the quarter. The funding plan for the AMD deployment is critical: by using asset-backed facilities, Rackspace avoids taking on additional unsecured debt. As Marino noted on the prior call, “it's going to be largely on par, if not accretive to existing gross margin rates across our Private Cloud business.” — Mark Marino, Chief Financial Officer · 2026-05-07 That sentiment is now being tested as the company scales the AI business. As Gajan had previewed in the prior quarter, “we are certainly well along the way there. I think we still need to get the financing locked down and sort of tightened up...” — Gajen Kandiah, Chief Executive Officer · 2026-05-07 Looking at the numbers, the gross margin trajectory shows a secular decline from over 40% in 2019 to 17.6% in the latest quarter, underscoring why the higher-margin AI services are so central to the rebuild. The company is effectively trading revenue for profitability, and the market has responded: the stock rallied 150% in the last 90 days before pulling back 56% from its June peak, reflecting both excitement and skepticism about execution.

The Bottom Line

Rackspace's AI pivot is both a company-unique strategic shift and a bet on a global theme—sovereign, inference-heavy AI infrastructure that many enterprises are just beginning to explore. The AMD agreement and Palantir partnership give it a credible slot in that narrative, and the Enterprise customers it already serves in healthcare, finance, and government provide a natural beachhead. The question is whether the capital-light financing holds together and whether the revenue-per-MW targets prove realistic. If they do, Rackspace could emerge as a niche consolidator in regulated AI—if not, it remains a high-volatility turnaround with a heavily levered balance sheet. This quarter, at least, the direction is clear.