Platformonomics TGIF #110: December 12, 2025

Oracle Disappoints: Q2 FY26

Oracle’s quarterly results were disappointing, and not just financially. Oracle is our gold standard for “bombastically scripted” earnings calls. Nobody does a shameless stock pump like the database vampire, but this one disappointed.

Admittedly they set a ridiculously high bar last quarter, which “was a bombastically scripted earnings call for the ages“, which resulted in perhaps the biggest one day stock pop ever (over $200 billion). But Oracle had already given all that back and then some before this week’s earnings.

They couldn’t match that last earnings call (or deliver on those promises…). Oracle stuck with the Remaining Performance Obligation narrative, despite the market already assigning a negative multiple to their previously announced RPO. (RPO is this era’s “eyeballs”, a non-GAAP metric supported in Oracle’s case by a non-binding “framework of a deal” that might or might not result in any future revenue).

Oracle “reported” RPO of $523 billion, up 438% from a year ago! Deferred revenue (the one you report to the SEC) was $3.28 billion, up 11% (and that includes SaaS).

It is bad to miss on revenue, both top line and for your AI cloud business, when your narrative is all about the future revenue you have booked.

Free cash flow continues its negative plunge (even with the sale of Oracle stake in Ampere). The debt/equity ratio is going to have to go up even more!

Oracle’s Remaining CAPEX Obligation: Q2 FY26

The flip side of Remaining Performance Obligation is Remaining CAPEX Obligation. How much do you have to invest to deliver the revenue? No infrastructure, no revenue. No CAPEX, no infrastructure. Oracle’s details on this topic are light, beyond pleading it won’t be as much as people fear.

Oracle spent $12 billion on CAPEX this quarter. They’re still upwards of $200 billion in cumulative infrastructure spend behind the HyperCAPEX companies, and fall further behind every quarter.

HyperCAPEX companies are sophisticated machines that efficiently turn huge amounts of money into cloud infracture. This machinery has been painfully developed over decades as those companies have climbed a steep learning curve. Oracle is early on this curve.

Today’s new data center was put into motion years ago. The HyperCAPEX manufacturing machinery is a multi-year process.

Which brings us to Oracle’s forecasting and what it says about the efficiency of their CAPEX machine. A quarter ago, Oracle projected CAPEX spending for the year would be $35 billion. Ninety days later, the projected number has grown to $50 billion (+43% in a quarter). You can always spend more to try to accelerate deployments, but eventually the pendulum will swing back to doing things efficiently.

Oracle also says things like this on their earnings call:

we don’t actually incur any expenses for these large data centers until they’re actually operational.

If that is true, Oracle’s already poor capital efficiency is deteriorating materially. Here we update our chart of how efficiently Oracle turns CAPEX into revenue (see Remaining CAPEX Obligation for the details and comparison to AWS, which provides the cleanest comp).

Which leads to the succinctly expressed question:

Oracle Still Can’t Build Data Centers: Q2 FY26

You may have heard this before, but the learning curve to deploy CAPEX at scale is very real.

Yelling at the contractor apparently didn’t help (and said contractor is trying to go public on the basis of this project…)

I have previously commended IBM’s analyst relations team as one of the rare pockets of out-performance at the company, but those days may be over, as new entrants like Semianalysis aren’t playing the old analyst game.

Read the whole thread.

Also worth reading the whole thread.

Supposedly McKinsey is now doing engagements where they get paid for results, not slides. We’ll see how that goes!

Clintonesque desperation from industry mascot army provider Salesforce? What possessive would Salesforce customers use?