Oracle: The Q2 EPS 'Decline' Is a Year-Ago Gain — Capital Intensity Is the Real Test

Oracle's fiscal first quarter, reported after the bell on September 10, was about as good as an AI-infrastructure quarter can look. Revenue rose 30% to $19.35 billion, adjusted earnings of $1.92 per share handily beat the $1.74 analysts expected, and the cloud-infrastructure business that has become Oracle's entire growth story more than doubled to $7.39 billion. Management raised the full-year bar to at least $90 billion of revenue and $8.10 of adjusted earnings per share. The reaction was telling for how far the stock has fallen: shares climbed in extended trading, yet the company entered the report down roughly 22% year to date, a huge gap versus a market that has gained.

That gap between the operating story and the stock is the whole controversy, and the confusing Q2 number in the headline sits right at the center of it.

The "flat" guide is a year-ago gain, not a slowdown

Here is the number that trips everyone up. For the fiscal second quarter, Oracle guided adjusted EPS of $1.85 to $1.93 in dollars ($1.83 to $1.91 in constant currency). Set that against the $2.26 Oracle earned in last year's fiscal Q2, and it looks like earnings are about to fall 15% to 18%. A reader skimming the headline would reasonably conclude the growth engine is stalling.

It is not. Fiscal Q2 of last year was inflated by a one-time pretax gain of about $2.7 billion from Oracle's sale of its interest in Ampere — part of a deliberate "chip neutral" data-center strategy. Management's guidance explicitly makes the point: strip out that year-ago gain and the new Q2 range is still 21% to 25% growth in dollars. The headline decline is an artifact of comparing this year against a quarter that contained a windfall, not a sign the business is rolling over.

The demand backing that growth is not a story, it is booked. Oracle ended the quarter with $664 billion in remaining performance obligations — contracted but not-yet-recognized revenue, up more than $200 billion from a year ago and above the roughly $630 billion analysts expected. The company signed over $30 billion of additional AI cloud contracts during the single quarter. That is the strongest forward-revenue visibility Oracle has ever printed, and it validates that the +121% IaaS line is real orders rather than promotional language.

The real trade-off lives on the balance sheet

So if demand is proven, why is the market so hostile? Because the growth is exceptionally expensive to deliver, and the Q2 guide inadvertently surfaces that cost.

Oracle spent $28.5 billion on capital expenditures in the quarter alone, versus $8.5 billion a year ago, adding about 850 megawatts of data-center capacity and more than 300,000 GPUs. Operating cash flow hit a record $23 billion, but that was not enough to cover the buildout: free cash flow was negative $5.4 billion for the quarter. To fund it, Oracle carries roughly $125 billion of borrowings, raised another $20 billion of common stock through an at-the-market program during the quarter, and saw interest expense jump 55%. Every AI contract Oracle signs is forcing it to build the capacity to honor it, which is why the RPO backlog and the capex bill are really two sides of the same coin.

There is a quieter consequence worth watching too. Oracle's GAAP EPS of $1.56 ran well below the $1.92 adjusted figure in Q1, and the gap — driven heavily by amortization — tends to widen as Oracle capitalizes billions of new data centers each quarter. So the adjusted earnings that investors anchor on will keep climbing, but the cash and the reported earnings underneath will lag, and the dilution that paid for the buildout keeps growing the share base.

Where the multiple sits after the fall

The valuation question is whether the market has already paid for all of that, and on this metric the answer leans yes. Oracle trades near $153, about 18 times EV/EBITDA — in line with Microsoft and arguably cheaper than Alphabet on the same basis, both of which produce positive free cash flow Oracle does not today. On the $8.10 full-year adjusted EPS guide, the stock is around 19 times forward earnings, with growth of 20%-plus still being guided. A name growing revenue 30% with a $664 billion backlog at that multiple is not priced as though the demand is in doubt.

The honest read separates the company from the stock. The company's thesis has been proven: this is monetizable AI demand flowing through to revenue, backlog, and a raised guide, not a narrative patch. But the bridge from here — whether negative free cash flow and relentless dilution convert into self-funding cash generation as the capacity comes online — is entirely unproven, and that is precisely the risk the market has spent a year repricing. Oracle is not obviously overpriced after the reset, but it is also not cheap enough to ignore the fact that the shareholder is funding the growth with debt and new stock. The next two to four quarters — when the freshly built capacity starts producing revenue against a rising amortization and interest bill — will decide whether the multiple was the discount or the trap.