Follow the CAPEX: The Clown Car Race Checkered Flag

We’ve used capital expenditures (CAPEX) to separate the clouds from the clowns for over a decade. Clouds consume copious CAPEX. Clowns confabulate.

It is time for another inspection tour of Clown City (previous visits: 2018, 2020, 2022). Once a bustling stop on the cloud-washing promotional circuit, Clown City is now but a shadow of its former self. Remember (colorfully bewigged) visitors like Cisco Cloud, HPE Helion, Rackspace, Verizon Cloud, and VMware Cloud? They all eventually learned – the hard way – that clouds are built with CAPEX, not chatter.

Sadly, a once-promising new generation of potential clowns never rose to the occasion of repopulating Clown City and entertaining us with the gap between their cloud rhetoric and their CAPEX spending.

The only activity left on the sleepy streets of Clown City is the CAPEX Clown Car Race between the last two of our self-proclaimed cloud pretenders: IBM and Oracle. It is time to declare a winner in this (admittedly bush league) race!

The CAPEX Clown Car Race checkered flag marks the end of the cloud-washing era. But clown fans should not despair. Clowns are eternal. The First Law of Clownodynamics says clownishness is neither created nor destroyed.

There are always companies cloaking themselves in the new hotness without possessing the proper prerequisites for participation. When the hotness moves, so do the clowns.

But some domains require real investment, which means the all-seeing-eye of CAPEX still exposes poseurs.

Hello AI clowns! Welcome to Platformonomics ClownWatch™

CAPEX might be an even better tell in AI than it was with cloud, because GPUs ain’t cheap and software now has COGS (a development which may prove regrettable).

But first, we have a drag race approaching the finish line.

The CAPEX Clown Car Race

Oracle got off to a very poor start in our race, with Larry Ellison harrumphing about cloud computing:

Maybe I’m an idiot, but I have no idea what anyone is talking about. What is it? It’s complete gibberish. It’s insane. When is this idiocy going to stop?”

Eventually Oracle got cloud religion after some embarrassing but certainly not unexpected attempts to substitute hot air for CAPEX. Nothing motivates like losing half your database market share.

Through 2022 it even looked like Oracle might be serious in the only way that matters: up-and-to-the-right CAPEX spending.

But 2023 saw a catastrophic 27% drop in Oracle’s CAPEX spending. While their absolute and directional spending is both disappointing and probably disastrous, it doesn’t matter because their opponent in the CAPEX Clown Car Race is IBM (company motto: “Down and to the right”).

There is a deeply philosophical question in how you win a clown car race. Is the winner the most clowny or the least clowny? With next to no consideration, we’re going with the latter and calling a winner.

Congratulations Oracle – based on your CAPEX, you’re the cloudiest of the clowns! But Oracle is still a clown, and certainly not a hyperscale cloud (there is no “big four” despite the best efforts of Oracle PR to slip away from the children’s table).

And IBM gets a clowniest clown participation trophy. IBM — I’ve always believed you could do this!

Below we’ll look at IBM and Oracle’s CAPEX spending, amuse ourselves with the current fairy tales they’re telling, ask what is going on at Oracle, and then make the brutal comparison with the CAPEX boomtown of Cloud City. And we’ll end with a quick “where are they now” look at that once-promising new generation of aspiring clowns.

IBM

Company motto: “Down and to the Right”

In what is not even remotely a surprise, IBM hit new record lows in 2023 for both absolute CAPEX spending and CAPEX spending as a percentage of revenue. The hypers of “hybrid cloud” cut CAPEX by 20%, spending less than $1.5 billion (across the entire company, not just cloud infrastructure), which is less than half of what they spent as recently as 2020.

These numbers confirm there is no cloud in “hybrid cloud”. Like soylent green, “hybrid cloud” is people – consultants (and mostly offshore consultants). There is no cloud business at IBM (as we have been saying for over a decade).

In perhaps related news, the leader of IBM Cloud quietly departed for another IT consulting company (one that doesn’t pretend to be a platform company).

But don’t shed too many tears for IBM. The company is in the midst of a bold strategic transformation. A transformation from a cloud 🤡 to an AI 🤡. Because the absence of CAPEX in either domain ensures you are a 🤡.

Meet the new shell game. Same as the old shell game.

IBM is now running its cloud playbook for AI. The company opened their Q4 2023 earnings release wanting you to think they are a generative AI company:

“In the fourth quarter, we grew revenue in all of our segments, driven by continued adoption of our hybrid cloud and AI offerings. Client demand for AI is accelerating and our book of business for watsonx and generative AI roughly doubled from the third to the fourth quarter,” said Arvind Krishna, IBM chairman and chief executive officer.

You’d almost think IBM was seeing material revenue from AI. But when asked to “elaborate specifically on exactly what the book of business means” by a veteran Wall Street analyst familiar with IBM’s antics, much bobbing and weaving ensues:

So Toni, on the AI book of business, this is not all revenue in the quarter. I would just begin with that statement to set it straight. At this stage, we wanted to start looking at what is our momentum, what is the sentiment from our clients. So we went to a measure that is more reflective of, I’ll use the word signings. What is the commitment the clients are making to us? Consulting is straightforward. It is the signings. Consulting signings are anywhere from 12 to 24 months on average is how much time they play out over there. And on software, it’s what they’re committing to. And we are using SaaS ACV. So it’s a 12-month commitment, which is typical for as a service as well as, since we do offer our portfolio both ways as license or as a service, it includes the license piece as well. Now over a long-term, let’s call it a couple of years or more, yes, the book of business should turn into an amount of revenue in a quarter, but that’s going to take a bit of time to catch up. But we felt that this gives the better indicator right now of what is our traction and what is our acceleration in that part of the business.

This is the same game IBM played for cloud. Chant the buzzwords and hope investors won’t notice the company isn’t actually investing and doesn’t have competitive products.

As they did with cloud, IBM’s AI numbers are not something they actually report to the SEC, yet lead with in investor communications. I don’t know if that is securities fraud, but I know IBM has no GPUs.

Oracle

Company motto: “That’s Not Just a Rack, It’s an Entire Region”

Data note: we do Oracle a solid by comparing their March 2023 through February 2024 (FYQ4 to FYQ3) results to the 2023 calendar year results of other companies. This lets Oracle bring a full two months from their glorious future to the comps. Hopefully that doesn’t give them too unfair an advantage.

When we last checked in on the database vampire, we noted they were (finally) ramping their CAPEX:

Oracle, on the other hand, is at long last starting to put its money where its mouth has been (though the mouth’s lead looks insurmountable). The database vampire doubled their annual CAPEX spend in the last year (using their March 2021 through February 2022 results). The company promised to spend $4 billion in CAPEX in FY22 and it looks like will hit that mark by May. And they surpassed IBM in annual CAPEX spending this year.

But after spending more than $8 billion in our ersatz calendar 2022, Oracle has fallen off the CAPEX wagon. Oracle CAPEX declined 27% to under $6 billion from 2022 to 2023.

Oracle’s commentary in the last several quarters has (commendably!) focused on CAPEX, with the CAPEX discussion taking up more time in each successive conference call. Oracle repeatedly tells us they have “enormous” cloud demand, are capacity constrained, and are building out data centers incredibly fast. But something has gone off the rails in Oracle’s cloud infrastructure build-out because their CAPEX spending has declined by double digits in each of the last three quarters.

In FY23 Q1 (Sept 2023), Oracle says:

Because we have far more demand than we can supply, our biggest challenge is building data centers as quickly as possible.

Capital expenditures were $8.3 billion over the last four quarters and we are clearly beginning to see the cash flow benefits stemming from our cloud transformation. CapEx was $1.3 billion in Q1 as we continue to build capacity for bookings and our customers’ growing needs.

Given the demand we have and see in the pipeline, I expect that fiscal year 2024 CapEx will be similar to this past year’s CapEx.

Yet Oracle CAPEX for FYQ1 was down 24%. And they reiterate the guidance to spend $8+ billion in FY24 CAPEX (ending May 2024).

In FY2024 Q2 (Dec 2023), Oracle told us:

The demand for Oracle’s Cloud Infrastructure and Generative AI is consistently increasing quarter after quarter.

We have to build 100 additional cloud data centers because there are billions of dollars more in contracted demand than we currently can supply. Cloud Infrastructure demand is huge and growing at an unprecedented rate.         

CapEx was $1.1 billion in Q2 as we continue to build capacity for bookings and our customers’ growing needs. Given the enormity of our pipeline and backlog, I expect CapEx will be somewhere around $8 billion this fiscal year, meaning our second half CapEx will be considerably higher as we bring online more capacity.

Yet Q2 CAPEX was down 56% and the lowest spend in eight quarters. And they slightly reduce guidance for the fiscal year ending in May 2024 to $8 billion.

By FY2024 Q3 CAPEX discussion dominated their conference call remarks (as it should in all calls):

Where if not for some continuing supply constraints, consumption growth would have been even higher.

Now while we spent $2.1 billion on CapEx this quarter, the $1.7 billion in the cash-flow statements is slightly lower just due to the timing of payments. So the $2.1 billion is actually what we spent and will pay for.

We are working as quickly as we can to get the cloud capacity built out given the enormity of our backlog and pipeline. I expect the CapEx will be somewhere around $7 billion to $7.5 billion this fiscal year, meaning our Q4 CapEx should be considerably higher.

We have enormous amounts of demand. I tried to make that clear last quarter, and we have more capacity coming online. But we have tried to – we’re trying to focus on much larger chunks of data center capacities and electricity and all of that and that’s just – that all to come.

Let me add that Oracle has been building data centers at a record level and a lot of people I think are aware that we can build fairly small data centers to get started when we want to.

We can go very small. We can get a full cloud data center with Ultra services in 10 racks.

But this is what I want to point out. We’re also building the largest data centers in the world that we know of. We’re building an AI data center in the United States where you could park eight Boeing 747s nose-to-tail in that one data center. So, we are building large numbers of data centers, and we were – and some of those data centers are smallish, but some of those data centers are the largest AI data centers in the world.

So, we’re bringing on enormous amounts of capacity over the next 24 months because the demand is so high, we need to do that to satisfy our existing set of customers.

So, for fiscal year ’25, looking at about $10 billion in CapEx because it’s also involves not only some big centers, but it also involves expansions of existing centers. So we’ve already got some areas that we will be filling out. So at least preliminarily, we’re looking at $10 billion for next year. And then it’s not too complicated to figure out the math here when I’m looking at somewhere between $7 billion and $7.5 billion for the full year and you’ve got all the numbers for one, two, and three at this point.

And I would include for Q3 the one we just are announcing. I would add in the amount we haven’t paid yet as the CapEx number for this quarter. Okay? And then I guess that would be and then Larry gets the second question. But anyway, so $2.1 billion for this quarter and you’ve got Q1 and Q2 and I’m going to be somewhere between $7 billion and $7.5 billion for the full year, which is actually a little bit lower than I thought. But we were able to do pretty well. You know-how we spend very carefully.

So, there is a tremendous amount of demand, the data centers take longer to build, and we would like that said, we are getting very good at building them quickly and getting the building the power and the communication links in, we’re doing faster than we have ever happened in the past.

It’s very different than it used to be. So, we’re able to bring additional capacity online very quickly if we have that the electric power and the communication lines. So, is the long pole in the tent is actually building the structure, connecting the electricity, connecting the communication lines.

Despite claiming they’re building data centers “at a record level”, Q3 CAPEX was down 36%. They strangely want credit for more CAPEX spending than what they reported to SEC (again, securities fraud questions are not addressed here). And Larry tried to distract with a soliloquy on how many 747s would fit in a future Oracle data center (when you only fly private, you may not realize airlines don’t fly 747s any more).

Guidance for the full fiscal year CAPEX spend has come down by over a billion dollars. The lower target still requires a huge ramp in Q4, and roughly doubling spending from Q3. But most of all, we see the first acknowledge of constraints on data centers build-outs, in particular electron availability.

And this double-digit CAPEX decline occurs amidst the AI frenzy, a trend Oracle, of course, claims to be leading. 2023 was an odd year to slash infrastructure investment if you’re big in AI (or say you are).

The fundamental question about Oracle’s cloud infrastructure ambitions is does the company just not know how to build data centers with more than a couple racks or have they concluded they don’t want to (or can’t) spend the kind of money this game requires?

Oracle has been on a very public cloud scale learning curve. First they pretended they didn’t need much CAPEX because, of course, Oracle’s stuff was just so much better than anyone else’s.

Then they won a fraction of Zoom’s cloud business, hyped it in time-honored Oracle fashion, and were humbled when they couldn’t fulfill the “extreme” capacity demanded.

While it is cute that Oracle counts a single rack as a data center and ten racks is a “Ultra” data center, the hyperclouds operate at many orders of magnitude greater scale. Oracle is now trying to transition beyond bonsai data centers and seems to be struggling with the concrete and steel world of data centers (with their need for building permits, connectivity and power).

No doubt Oracle will someday learn why hyperclouds are building smaller data centers and not 747 hangars.

The hyperclouds have decades of experience building cloud-scale data centers at this point. They manage multi-year lead times and are deeply engaged in the “mad scramble” for power, which is the biggest constraint for data centers today.

Oracle is now guiding to a FY25 CAPEX spend of – gasp — $10 billion, but given their recent inability to actually spend to their guidance, that may still be optimistic.

The other possibility is Oracle is having second thoughts about the financial hit associated with transforming an asset-light software company into a serious cloud infrastructure player. In 2022, Oracle’s CAPEX as a percentage of revenue hit 18%, comparable to what hyperclouds can spend. But that spend also came with a roughly $2 billion profit hit. They quickly reined in that percentage by over a third in 2023.

But the financial issue isn’t just the impact on Oracle profits. Tragically, they aren’t actually competing with fellow clown car IBM. They’re competing with Amazon, Google and Microsoft and must soberly contemplate their financial wherewithal to play in that league.

For all of Oracle’s ramp in CAPEX in recent years, they are over $150 billion behind cumulative hypercloud spending. And they are falling further and further behind with every passing quarter. In 2023, AWS spent $24.8 billion on CAPEX (over four times Oracle), Google spent $32.3 billion (over five times Oracle), and Microsoft $41.2 billion (almost seven times Oracle). Microsoft spent more on CAPEX in just 2023 than Oracle has in its entire nearly 50 year history.

Oracle’s $10 billion in CAPEX guidance for FY25 adds a nice incremental zero to their spend (and hits a milestone the hyperclouds all passed in 2016), but it still doesn’t come close to what the hyperclouds are spending. If you add up Oracle’s actual CAPEX spending and guidance for their fiscal years 2022 through 2025, it doesn’t match what Google and Microsoft spent in 2023.

To give a sense of how dramatically Oracle is being left behind, here is the combined CAPEX spending of AWS, Google and Microsoft (corporate level spend for Google and Microsoft, but ~80% of that is cloud infrastructure), minus Oracle’s spending (which is also corporate level). The gap has grown from $20 billion in 2015 to over $90 billion in 2023.

For completeness, to make Oracle’s clown car status abundantly clear, and to make my biennial favorite joke about Oracle being indistinguishable from the x-axis, here is annual and cumulative CAPEX for both the hyperclouds and Oracle.

The Oracle plan to catch up to the hyperclouds remains unclear to me.

What Ever Happened to…

While our once-promising cohort of new clowns disappointed in both their hyperbole and investment, they remain entertaining in other ways:

• Bank of America – still dominant as the worldwide leader in financial crime, but is also now insolvent in the wake of rising interest rates. And still have spent ZERO on CAPEX since 2011 (bank accounting FTW). Their alum and pal at IBM who was leading the charge for a compliance-oriented private cloud doohickey bailed, so expect B of A to eventually be the last bank to embrace the public cloud.
• GAIA-X – the EU’s hand waving assertion of their strategic cloud autonomy “based on European values” is, in the words of one booster, “a paper monster that will exist but will not have any impact in the market”. Others say it “failed to reach critical mass or really take off”, “operates too much at the meta level”, is “still a long way from clear, tangible use cases” and “never got out of this theoretical sinkhole”. Even IBM spends more on CAPEX in a year than GAIA-X’s multi-year budget.
• Trump Media & Technology Group – “Net cash used in investing activities for the twelve months ended December 31, 2023 was approximately $2.2 thousand” ($2,200.00). Emphatically serverless!

We now return to our regular programming.