Rackspace Jumped 13% on an NVIDIA Deal. The Balance Sheet Still Owes $2.8 Billion.

On the morning of September 10, Rackspace Technology (NASDAQ: RXT) joined the NVIDIA Cloud Partner Program and unveiled an "Institutional Sovereign Pod" — NVIDIA Blackwell compute paired with Palantir's Foundry and AIP, wrapped in a governed private cloud built for regulated enterprises and governments. The stock jumped more than 10% on the news. Headlines read it as a validation event, another tier-2 infrastructure name hitching itself to the AI trade.

It is worth reading the press release more literally. Rackspace's role in the collaboration is not to build chips or write models; it is to be the "accountable operator" — the company that actually runs the whole system, from the GPU to the governed data layer, where uptime, sovereignty, and compliance are non-negotiable. CEO Gajen Kandiah put the pitch plainly: as enterprise AI becomes production infrastructure, the question shifts from "who provides the GPU" to "who is accountable for operating the complete system." That is a real layer, and arguably the right one. NVIDIA is making compute abundant; what becomes scarce in response is trust — an operator a government or a bank will let near its most sensitive data.

The problem is that a 13% jump in Rackspace's equity is not a statement about the sovereign-AI opportunity. It is a statement about Rackspace's balance sheet. And the balance sheet is the story the headline buried.

The stock moves like a call option, because it is one

Look at the most recent quarter. Rackspace reported second-quarter 2026 revenue of $670.1 million, up 0.6% from a year earlier. Private Cloud grew 5.5%. Public Cloud — the resale-and-migration business — fell 2.3%. The company still lost money: a GAAP net loss of $67.5 million for the quarter.

Now the part that matters. As of June 30, 2026, Rackspace carried about $2.79 billion in total debt against $111 million in cash. Stockholders' equity was negative $1.28 billion. Total liquidity, including the revolving credit line, was $202 million. Relative to last-twelve-month EBITDA of roughly $287 million, net leverage sits around 8x — nearly double the roughly 4x leverage Apollo loaded the company with when it took it private for $4.3 billion in 2016 and put $3.2 billion of that deal's debt on the company's own books.

That is the mechanical reason a partnership announcement moves the equity so violently. There is almost no book equity beneath the stock. When a deeply levered company catches good news, the equity is a small cap on a large pile of debt, so a modest improvement in the enterprise's perceived prospects swings the equity by a large percentage. The same mechanism explains the downside: Rackspace's shares traded between $0.39 and $8.60 over the past year, with a five-year beta above 3. The unsecured bonds due in December 2028 trade at 84 cents on the dollar — the debt market's own estimate that this is a distressed credit, not a clean new-tech growth story. The equity is effectively a call option on whether the AI buildout succeeds in time.

The recent record says the AI pivot is not free

The bullish read treats the NVIDIA and AMD partnerships as pure upside. Two months ago the evidence pointed the other way. On July 9, Rackspace cut its full-year 2026 revenue guidance by $150 million and disclosed that its AI investments required a significant re-prioritization of resources, including a "transition away" from revenue-generating segments. The stock fell 33.6% in a single day, from $6.58 to $4.37.

A securities fraud class action, with a lead-plaintiff deadline of September 28, 2026, alleges the company overstated the financial impact of its AI push during a class period running May 7 through July 8 — that the AI buildout pulled capacity and capital away from the profitable Private Cloud segment even as management reaffirmed its growth profile. Allegations are not findings, and this is a heavily litigated corner of the market. But the timing is the point: Rackspace's AI story produced a guidance cut, not a guidance raise, barely sixty days before Thursday's NVIDIA announcement.

The money question is financing. Rackspace's AMD partnership targets a 30-megawatt footprint by 2028, which management estimates could eventually deliver $450–600 million in annual revenue and, at roughly 50% margins, $225–300 million in EBITDA. The buildout is estimated to need around $900 million in total capex. The company financed it with a $250 million at-the-market equity offering — which, by definition, dilutes the same small float that the +13% day just pumped up. It covers a fraction of the deployment cost, and the $2.5 billion debt load matures in 2028, the same year Rackspace targets full deployment of that AI infrastructure.

The real question the headline dodges

Here is the clean split, and it is the discipline of this entire story. The sovereign-AI thesis is genuine — a governed operator for regulated and sovereign workloads is a defensible position in a market where the hyperscalers hoard capacity and governments demand data residency. The partnership announcements are real. But Rackspace, the specific instrument, is not a clean expression of that theme. It is a leveraged turnaround whose AI upside has to pay for eight-plus turns of debt and a negative-equity balance sheet before it does anything for shareholders — and the equity itself is being diluted to fund the buildout.

The data that would change the picture is specific and observable: firm GPU revenue commitments showing up in the numbers rather than in press releases, financing that closes the gap between the $250 million raised and the roughly $900 million needed, and a guidance trajectory that stops contracting. What would break the bullish case is another cut — another day when the AI pivot, which is real and strategically sensible, once again requires a "re-prioritization" that lands on the revenue side of the income statement.

For now, the +13% has the mechanism exactly backwards. It treats the partnership as validation of the company. The balance sheet says the company is validation of nothing — it is a highly leveraged option on execution, and the option is priced in cents. When the headlines write themselves, read the debt first.