A year ago, Oracle's contracted cloud backlog stood at $455 billion, itself a record at the time. On September 10, 2026, the company reported fiscal first-quarter results showing that backlog has grown to $664 billion, comfortably ahead of the roughly $640 billion Wall Street analysts had projected. Oracle shares rose 6% to 7% on the news, a real vote of confidence from a stock that, despite the jump, remains down 22% year to date against a broader market up around 11%.

The quarter's headline numbers were strong across the board. Total revenue reached $19.3 billion, up 30% year over year, and adjusted earnings per share came in at $1.92, ahead of the $1.74 analysts expected. But the number that mattered most to investors was Oracle Cloud Infrastructure revenue specifically, which more than doubled, climbing 121% to $7.4 billion.

A Number That Beat Wall Street by $24 Billion

Oracle booked more than $30 billion in new AI cloud contracts during the quarter alone, the single largest driver behind the backlog's jump from $455 billion to $664 billion. The company said roughly half of that total backlog is expected to convert into actual recognized revenue over the next 36 months, a detail that matters because a contracted backlog only means something if a company can actually deliver the infrastructure those contracts promise.

Oracle's own language in its earnings disclosure was direct about the underlying dynamic: customer demand for AI cloud training and inferencing services continues to grow faster than supply. That's the same capacity-constrained story that's defined Microsoft's own record Azure quarter and Amazon's admission that AWS has already sold more cloud capacity than it can currently build, now showing up just as clearly in Oracle's own numbers.

Where the Growth Is Actually Concentrated

Oracle has spent the past two years transforming its public image from an enterprise software veteran into a genuine AI infrastructure heavyweight, and this quarter's numbers back that shift up with real capacity metrics. The company delivered 850 megawatts of additional data center capacity during the quarter and has shipped more than 300,000 GPUs to AI cloud customers since the end of its fourth quarter, almost triple the pace of capacity it was delivering previously. GPU utilization across Oracle's fleet reached 97.9%, a figure that leaves almost no idle capacity sitting unused, and underscores just how tightly stretched Oracle's infrastructure is against current demand.

One detail addresses a concern that's followed Oracle's AI story for over a year: how dependent its growth actually is on a single customer, OpenAI. According to SiliconANGLE's analysis, Oracle's non-OpenAI backlog has more than doubled over the past year, evidence the company's AI infrastructure business is broadening across a wider customer base rather than resting entirely on one enormous, concentrated relationship.

The Real Question Isn't Demand, It's Delivery

Oracle's software business, the company's traditional core, told a less flattering story this quarter. Software revenue reached $5.55 billion, down 3% from a year earlier and below the Street's $5.61 billion target, a soft spot that got partially offset by a fresh win: a new Pentagon contract worth up to $7 billion over the next decade, covering Oracle software deployed in on-premises data centers used by the US military, the intelligence community, and the Coast Guard.

Analysts covering the report increasingly frame Oracle's real challenge as execution rather than sales. As one analysis from TradingKey put it, the question is no longer whether Oracle can land major AI contracts, but whether it can build capacity fast enough to actually fulfill them. Delays in securing power, permitting, construction, or skilled labor could all slow how quickly that $664 billion backlog turns into recognized revenue, regardless of how impressive the contract figure looks on paper today.

How Oracle Is Paying for All of This

Building capacity at this pace carries a steep near-term cost. Capital expenditures reached $28.5 billion for the quarter, up sharply from $8.5 billion a year earlier, and free cash flow turned negative, landing somewhere between $5 billion and $5.4 billion in the red, against a debt load CNBC estimated at roughly $125 billion. To help fund the buildout, Oracle also completed a $20 billion gross at-the-market equity offering during the quarter, gradually selling new shares into the open market to raise fresh capital, a similar financing pattern to what other capital-intensive AI infrastructure builders like Crusoe have used to fund their own rapid data center expansion.

Notably, Oracle said a meaningful share of its capital spending, roughly $11.36 billion this quarter, was offset by customer prepayments, with some newer contracts structured around customers pre-funding infrastructure or supplying their own hardware. CFO Hilary Maxson told analysts that despite the scale of new contracts booked, Oracle's full-year capital spending guidance and data center delivery timelines remain unchanged, and that the structure of these newer deals means they don't add incremental pressure to the company's broader capital-raising plans.

A Tale of Two Earnings Weeks

Oracle's results landed just three days after software company Adobe issued a weaker-than-expected sales forecast, warning that AI's disruption of traditional software workflows was weighing on its own business. Startup Fortune drew a sharp line between the two reports: Oracle's backlog grew because AI buyers still need chips, data centers, and power, while Adobe's stock reaction reflected the harder reality that selling software applications built on top of AI has proven considerably tougher than selling the raw infrastructure AI runs on.

What $664 Billion Actually Confirms About the AI Trade

Oracle isn't the only company reporting numbers at this scale this week. Microsoft has mapped out a 38-gigawatt data center buildout, and the Pentagon has separately opened discussions around a $5 billion loan program specifically aimed at shoring up strained US data-center supply chains. Taken together, Oracle's $664 billion backlog isn't really a story about one company's specific execution, impressive as its recent delivery pace has been. It's further confirmation that the infrastructure layer underneath AI, the physical chips, power, and data centers making all of this possible, remains the part of the industry where demand is least in question. The part still genuinely uncertain, for Oracle and every company chasing this same buildout, is whether the physical world of power grids, construction permits, and chip supply chains can actually keep pace with backlogs this large before something in that chain buckles under the strain.