The reaction that made no sense
Oracle (ORCL) reported strong earnings, but Wall Street disliked the stock. Shares jumped about seven or eight percent the day the results came out. They traded well for the first couple of hours the next morning, then fell off, ending down about 1%.
The results were good in the exact places where there might have been concerns. The main concern was growth at Oracle Cloud Infrastructure (OCI), the data center business. That business builds a data center for a customer, the customer moves their operations in, and only then does Oracle start getting paid. Oracle spends billions - possibly tens or hundreds of billions when it is all done - to build these centers before any payment comes back. The results show they are now getting paid: OCI revenue rose 121% year over year, even while Oracle is still deep in the buildout.
What the numbers showed
Customers are paying for the work of building data centers, and they are also buying more software - ERP and database software. Oracle services revenue fell about 1% year over year. Both total revenue and the growth rate itself rose, at the data centers and across the whole company. A strong quarter.
The problem is likely a translation issue - Oracle trying to tell its story to investors. Expect more of that story at Oracle AI World in Las Vegas in October, with the analyst day near the end. Ken Bond, Oracle's longtime and well-liked head of analyst relations, announced his retirement on the conference call. That is not what is moving the share price.
Oracle (ORCL) has become the place where AI doubt collects. When other companies announce big AI contracts or work with the biggest players, their stocks rise. When Oracle announces contract wins, the market does not like it. Oracle clearly has enough OCI business from OpenAI, and many other companies are now contracting with Oracle to build these centers.
Hard projects on purpose
Co-CEO Clay Magouyrk said on the call that they expect these projects to be hard and to hit difficulties. If it were easy, everyone would do it. If everything goes on schedule, the schedule was not tough enough. Oracle appears to be responding to community concerns at project Jupiter in New Mexico, its large Abilene facility, and its Wisconsin site, where those concerns differ by location. That suggests they will finish these projects, and the numbers back it up. Earnings matter because press releases can say anything between reports; the earnings results carry real numbers. The 10-Q usually comes at the end of the trading day after earnings.
Higher rates and cash concerns
Could higher yields be the worry? The data center buildout is largely immune to higher rates, according to commercial real estate people. So the concern shifts to Oracle's cash position versus other hyperscalers, its credit rating, and negative free cash flow.
It is pedal to the metal for the AI buildout right now. Oracle recognizes this is the moment to build; if they do not do it now, someone else will. So they are willing to take the hit to free cash flow. Owning Oracle shares is a bet that OpenAI will pay its bills, that AI will matter in the future, and that these data centers will get built. If you do not believe those things, short the stock. My view: OpenAI will raise the money to pay its bills and may build a profitable business, and Oracle will get paid for these projects.
The 20% premium on old chips
A brief but important point from the call: customers re-signing with OCI are paying a 20% premium over what they paid before, and they are using the old chips. Almost a year ago, hidden in Oracle's filings, the company extended the useful life of its data center buildings, computers, and chips - one of its revenue recognition segments - to six years instead of five. That suggests the chips are still good six years later. That is hard to imagine, because every year Nvidia (NVDA) releases a new chip and you want the new one, not the last one.
But Oracle said customers are paying 20% more to use the old Nvidia chips. That changes the math when you discount financial models and estimate costs four, five, six, or ten years out. When Oracle has a 20-year lease on a building and a 5-year deal with a customer, the natural worry is what happens in year six. These results show customers re-signing at higher rates to use old chips. That means the bet on data centers has already been a great one. This is very positive. My biggest worry is that all the other Wall Street analysts also like these results - being on the same side as people better at this than me makes me nervous.
Analyst moves match a rerating, not a turn bearish. RBC lowered its price target but kept sector perform. Stifel lowered its target but kept a buy rating and cited the strong quarter. Analysts are playing catch-up.
Beyond data centers: agents
Oracle loves telling customer stories and keeps reference accounts, sending a separate press release each quarter with customer wins. The layoffs and workforce changes have been watched closely for whether Oracle is seeing efficiency and productivity gains. Oracle is talking about AI agents for human resources teams and mentioned a Pentagon contract.
The focus will shift from data center buildout toward agents and what agents can do. When you build financial models and try to imagine how many users a product could have, and the users become agents, the use cases become infinite - so spending can become infinite. Growth could be far larger than any group of humans could produce, and that is showing up in the results. Many Oracle software changes make it easier for users to create agents. That drives use of Oracle software, drives use of data centers, drives use of AI models, which drives more data center use, which drives Oracle's revenue. This is not a prediction; it appeared in the results.
Larry Ellison, chief technology officer, and the new co-CEOs Mike Sicilia and Clay Magouyrk see this as the moment to spend and get in position. It is a different business from the software business that carried Oracle for decades. The conference calls now lack Larry Ellison going off to left field as he did for years. This is a new era for Oracle, built on a new data center unit, which is why so much attention is on that business rather than the strong software business.
