Digging into the bottlenecks of AI

**Tyler Crowe** (0:02)
Diving deep into the AI Supply Chain today on Motley Fool Hidden Gems Investing.
Welcome to Motley Fool Hidden Gems Investing. I'm your host, Tyler Crowe, and today I'm joined by long-time Fool contributors, Jon Quast and Matt Frankel. So we wanted to start the discussion today about AI spending basically, and some of the nooks and crannies that a lot of that is falling into. And to do that, we want to start with the big news event of the day, and that was Oracle's earnings. They reported earnings early today, and the stock is down. Pre-market was down about 11 percent right around the time that we are taping. And the numbers were pretty good, at least from a quarterly earnings perspective. But I think the biggest reason we saw this massive decline in the shares was because CAPEX spending is going way, way up. They spent much more than they had expected. They were expecting $50 billion for this entire fiscal 2026, but they ended up spending 55
And their plan for 2027 was to add more to that, do about $70 billion in capital spending. Now, there's a lot of reasons for that. The remaining performance obligations, basically however hyperscalers are spending has gone way up. And we'll get into that. So guys, one of the things I think about with Oracle compared to like the other hyperscalers is, this is not like their bread and butter business. This is something they've been really getting into. And is this, seeing these kind of large, large numbers, is this like a risky approach? Is this like the way Oracle should be attacking this?

**Matt Frankel** (1:35)
You're right that the earnings report was very strong, but the earnings are not the full story. If they were, the stock would probably be up today.
Both revenue and earnings grew by 20% or more. They beat expectations. Cloud revenue grew by 47%.
The RPO number that you mentioned, it is simultaneously the company's biggest risk factor and the biggest bull case. Really, having a $638 billion backlog would be a dream of most companies. But when it's all tied to one customer, it does seem to be diversifying a little bit. It's now estimated that over 50% of that backlog comes from OpenAI as opposed to virtually all of it not that long ago. The number that you just mentioned, it increased $85 billion sequentially.
That's not OpenAI deals. Prepaid AI contracts, it's worth mentioning now, total $75 billion of that. That kind of helps the need to raise capital, but there's still a pretty big gap there. Having said all that, the company does have a massive capex need. $56 billion in fiscal 2026 That's up from $21 billion a year ago. Negative free cash flow for the first time in a long time and probably for the foreseeable future, if we're being honest. It could certainly pay off if OpenAI and the others can afford to pay for all of their committed spending, but it's still a risk factor in the near term. Almost $130 billion in debt and that's growing fast. And that anytime you're adding debt quickly, you're adding risk.

**Jon Quast** (3:00)
To your point, Matt, there's really no other way to do this, to be fair. When the space is moving as fast as it is and you've got to spend a ton of money to build something that's very expensive up front, of course you're gonna take on a lot of debt and you're going to take it on very quickly. So I want to be fair to Oracle on the one hand and also I want to acknowledge that it is a long game that the company is playing. This is, to the RPO number specifically, okay, $638 billion, yes, but over half of that is more than three years out. So it's not expecting this money anytime soon. Now, that's a little bit risky for sure.
Do I think that AI data centers are gonna still be a big thing in three years? Yes. Do I think that all of that money is necessarily in the bag right now for Oracle? No, not necessarily.
That said, I mean, OpenAI, much of what it is doing is the Stargate Initiative, and there's a lot of companies involved in that. Also, the White House is supportive. So it's risky, but I wouldn't say that it's, you know, humpty-dumpty risky.

**Tyler Crowe** (4:11)
To your point with the idea of it being risky or not or something like that, one of the things that is, it does seem like these companies are really dependent on, Matt, you said it was over 50 percent of it is at OpenAI. I'm assuming a lot of it is Anthropic. We've seen Anthropics, OpenAI, these other AI companies in the world, they have been signing some very large checks related to spending on compute and stuff like that. Part of the reason we, when we discussed the S1 for SpaceX, I think it was last week or a couple weeks ago, we were talking about Anthropic doing, I think it was $1.2 billion per month in basically renting compute from SpaceX, and a lot of what we're seeing here with these RPOs are similar deals. These are big, big numbers and a lot of money that they have to even cough up upfront. That gets me to one of the things I've also seen. OpenAI just confidentially submitted their S1, so they're also planning on going public relatively soon. Anthropic already did it. We're expecting that S1.