A quick clown CAPEX post since the Platformonomics TGIF weekly newsletter is done for the year.
UPDATE: I got the spend number wrong. Oracle tries to make their CAPEX number look bigger by showing YTD or TTM numbers as opposed to the quarterly number. Despite making fun of them for this, I fell for it this quarter in haste to write up. Edits below. Their spend is still decidedly not hyperscale.
Oracle (company slogan: “that’s not just a rack, it’s an entire region!”) announced CAPEX spending today (earnings too probably).
Oracle spent $2.89 $3.89 billion on CAPEX. That is record quarterly spend for the database vampire.
But they still face two problems on the CAPEX front.
First, Oracle is struggling to meet their CAPEX guidance from a quarter ago (which was $13.7B for the fiscal year, double the previous year), continuing the company’s pattern of being unable to reliably predict CAPEX spend even a quarter out. With $2.3B in CAPEX in the first quarter, Oracle needed to average $3.8B per quarter to hit the annual guidance. With Q2 almost a billion dollars below that (-24%), their year is now heavily back-weighted, needing to spend $4.25B in each of the final two quarters to hit their (decidedly neocloud-scale, not hyperscale) guidance. That is a big leap for a company that has only recently stopped outfitting their data centers at Best Buy, and has struggled to show they can systematically convert cash into compute infrastructure at scale.
Second, after a late start and spotting the hyperclouds a ~$200 billion lead in CAPEX spending, Oracle falls further and further behind with every passing quarter. Membership in the hypercloud club requires more than just braggadocio – it required beaucoup CAPEX.
(Amazon is not shown because we don’t have quarterly data for AWS CAPEX, only annual, but it tends to be a little less than half of Amazon’s overall CAPEX, But kudos to all the financial analysts attributing all of Amazon’s CAPEX to data center spend!).
After a number of quarters with colorful commentary (“gosh, it turns out it is hard to find power”) chronicling Oracle’s gradual ascent up the CAPEX learning curve, the company has gone silent in this most recent quarter. They reiterated their $13.7B CAPEX guidance for the fiscal year and that is it about it.
And there was no follow up on Larry’s Seemingly Unscripted Moment™ from last quarter about maybe deploying small modular reactors some time, somewhere. Meanwhile, the real hyperclouds have all made more substantive nuclear power announcements (see our Cloud Reactor Tracker). Larry did claim, in this quarter’s Seemingly Unscripted Moment™, Oracle is building a 1.6 gigawatt data center. But in the absence of any power announcements and their miniscule CAPEX outlays, it is hard to believe that shows up this decade (if ever).
Previous:
Oracle’s Data Center Difficulties: FY25 Q1
Oracle’s Data Center Difficulties
Oracle Still Can’t Build Data Centers
Why Can’t Oracle Build Data Centers?
