Rackspace Told Investors the AI Story. The Stock Funded It.

Rackspace announced a partnership with AMD on May 7, 2026. The same day, the company reaffirmed its full-year revenue guidance. The stock had already nearly quadrupled year-to-date on the AI infrastructure narrative. Two months later, Rackspace revealed that the AI strategy it had been building that narrative around would require pulling capacity and capital away from its profitable Private Cloud business. It cut its 2026 revenue guidance by $150 million. And on that very same day, July 9, it launched a $250 million at-the-market equity offering - described by the CFO as "100% primary," meaning fresh capital for the company, not insider selling - to fund the GPU expansion that was cannibalizing the existing revenue engine.

The stock fell 33.6% in one session, from $6.58 to $4.37. A securities fraud class action followed. The lead plaintiff deadline is September 28, 2026. The case is captioned Morgan-Reed v. Rackspace Technology, Inc., pending in the Southern District of New York.

The law-firm press releases about this lawsuit are identical boilerplate: buy between May 7 and July 8, contact us, there is no cost to you. But the actual structure of what happened here is more interesting than the litigation template suggests. The weird part is not just that the guidance came down. The weird part is that on May 7, Rackspace told investors the AMD deal "give[s] [the Company] confidence in the full year Private Cloud growth profile", while simultaneously beginning a capital reallocation plan that would strip Private Cloud of the capacity that makes it profitable. That is the disclosure boundary the lawsuit turns on. And it is the boundary that determines whether this is a painful but legitimate strategy shift or something closer to misrepresentation.

Let me compress the timeline because the class period itself tells the story. It opens on May 7, the day Rackspace signed the AMD memorandum of understanding and reaffirmed guidance. It closes on July 8, the last trading day before the earnings bomb. During those two months, the stock kept climbing on the narrative that Rackspace was becoming "the operator of the full enterprise AI stack," with partnerships across AMD, Palantir, VMware, Rubrik, and Uniphore. On the Q1 earnings call in May, CEO Gajen Kandiah said the market was moving in the direction Rackspace anticipated. He sounded like someone executing a plan, not discovering one.

Then on July 9, the plan met the balance sheet. Rackspace disclosed that AI investments would require a "significant re-prioritization of resources" and a "transition away" from certain revenue-generating segments. The revised full-year revenue guidance: $2.45 billion to $2.55 billion, down from the $2.6 billion to $2.7 billion range given in May. Private Cloud revenue was cut by $25 million. The company said lower near-term margins reflect "upfront growth investment and restructuring, ahead of AI revenue ramping."

The simplest model here is that Rackspace is a capacity-arbitrage business. It owns data center infrastructure and sells cloud hosting - some of it privately (Private Cloud), some as a reseller of hyperscale platforms (Public Cloud). The profitable part is Private Cloud. The part that was already leaking was Public Cloud, where customers were contracting directly with AWS, Azure, and Google, bypassing the Rackspace resale layer. The company knew this was happening. The complaint alleges it did not disclose it. And the new AI strategy requires buying expensive GPUs and putting them in data centers that could otherwise be running Private Cloud workloads for paying customers.

That is the basic tradeoff: the AI pivot looks forward, but it funds itself by pulling capacity backward. You cannot run both at full tilt with the same infrastructure. Rackspace apparently chose AI, then spent two months telling the market it would not have to give up Private Cloud while it figured out how to pay for the GPUs.

The $250 million ATM offering is where the incentive structure gets clearer. An at-the-market program lets a company sell shares continuously into the market through a broker-dealer, at whatever price the stock happens to be trading at. The CFO said the proceeds would fund GPU growth. The timing - launched the same morning as the guidance cut - means the company needed capital at the exact moment it revealed its near-term earnings would crater. That is a rational move: raise equity before the market fully prices in the hit. But it also means the $250 million is dilutive for existing shareholders, and it dilutes them at a price built partly on the Private Cloud growth story the company would then walk back.

The lawsuit alleges five specific omissions. In order: (1) AI efforts would require pulling capital and capacity away from Private Cloud; (2) Public Cloud revenue was already declining as customers went direct with hyperscalers; (3) Rackspace would need to reduce a material portion of its Public Cloud resale business; (4) all of this would meaningfully impact fiscal 2026 revenue; and (5) positive statements during the class period therefore lacked a reasonable basis. The complaint does not allege that Rackspace's AI strategy is bad. It alleges that management told investors Private Cloud was safe while internally planning to strip it.

There is a real counterargument here. Rackspace's AI margin economics are actually compelling, at least on paper: management expects revenue of $15 million to $20 million per megawatt, with EBITDA margins above 50%. The company outlined a multiyear plan targeting 30 megawatts by end-2028. If that ramps, the Private Cloud hit is a transitional cost, not a structural impairment. Management could argue that in May, they genuinely did have confidence in Private Cloud because the cannibalization had not crystallized yet - you can hold a plan at the strategic level without having run the capacity math down to the quarterly guidance number. The "re-prioritization" was real, but the timing and magnitude of the revenue impact may not have been known with certainty. That is a defense worth litigating.

But the May 7 language - "confidence in the full year Private Cloud growth profile" - cuts in the other direction. That is not vague forward-looking commentary. It is a direct statement about the profitability of the segment the AI pivot would hurt. If management knew then that capacity was going to move, the statement looks less like optimism and more like a bridge between the AI narrative and the guidance number that needed to stay intact.

The stock is at roughly $4.92 today, having recovered some from the $4.37 trough but still well below where it was before the July 9 earnings. Over the prior 120 days, it had gained more than 1,000%. Year-to-date, it was up over 400%. The kind of run that makes an ATM offering look like a rational capital-raising window, and also makes a 33% single-day drop look like a violent repricing of everything the narrative promised.

The class action is the standard machinery: investors who bought between May 7 and July 8 and lost money can seek lead plaintiff status by September 28, or do nothing and remain absent class members. No cost, contingency fees, court-approved, the usual plumbing. Whether the case survives a motion to dismiss depends on whether the court finds that the May statements were objectively misleading at the time they were made, given what management knew. That is a fact-intensive question about internal planning versus public communication, and it will turn on documents we do not have access to.

The more structural point, and the one that matters for anyone still holding RXT shares, is that the lawsuit is just the legal label on a capital-allocation problem. Rackspace is a company that was good at one thing - managing and reselling cloud infrastructure - and is now trying to become good at something else - operating enterprise AI compute - while winding down the first business that actually generated predictable cash. The AI margins look better in theory. The question is whether the company can finance the transition without destroying the earnings base that holds the stock price up during the build. The $250 million ATM was one attempt to answer that. The guidance cut was the admission that it was not enough on its own.

The lawsuit does not change the business. It just formalizes the gap between what management said in May and what the balance sheet showed in July. Whether that gap is fraud, bad planning, or a legitimate shift that got communicated too optimistically is what the next year of discovery will sort out. But the machine itself - sell the narrative, raise the equity, reveal the cost - is old finance in an AI costume.