Is Oracle a Construction Company?

The $185 Billion Lease Agreement

Ordinarily, when a very large company holds a big meeting and announces a plan to more than triple its revenue in five years, its stock goes up. It is a sign of confidence, a vision of explosive growth. It is the kind of thing that makes Wall Street happy.

And so it was a little weird last week when Oracle, the 48-year-old software giant, did exactly that, and its stock... slipped. The company projected its annual revenue would hit a staggering $185 billion by 2029—a massive leap from the $57 billion it just reported for fiscal 2025—and a 78% jump from the already ambitious target it set just last year. But instead of cheering, the market seemed to yawn and ask a simple, nagging question: That's nice, but who's paying for it?

One way to think about this is that Oracle has transformed itself into a 'high-stakes landlord' for the AI boom. It's no longer just selling software; it's building custom, hyper-specialized 'AI Super-Factories' for a new generation of tenants. OpenAI is the star tenant, having signed a massive, multi-decade lease agreement (worth a reported $300 billion). On Thursday, Oracle announced it has signed up even more tenants, like Meta.

The problem is, to be a landlord, you first have to build the building. And Oracle didn't say much about the construction costs. As Dan Gallagher at The Wall Street Journal pointed out, there's a rather large hole in the balance sheet:

What Oracle didn't say is how it expects to pay for the very expensive expansion of its network that will be needed to generate such returns... The company's capital expenditures exceeded its operating cash flow for the first time since 1990 in its latest fiscal year.

That is likely just the start of it. Wall Street expects Oracle’s negative free cash flow to continue for the next three fiscal years, with cash burn for the period totaling nearly $29 billion by the end of fiscal 2028...

This is the hard part. Before Oracle can collect a single dollar of that rent from OpenAI, it has to first go out and spend tens of billions of dollars in real, hard cash on construction—buying land, pouring concrete, and filling its factories with pricey Nvidia chips. Wall Street now expects Oracle to burn through nearly $29 billion in cash over the next three years to fund this build-out.

To pay for it all, Oracle is taking out massive loans. The company just sold $18 billion in bonds, and according to Morgan Stanley's debt analysts, that will only cover about a quarter of its cash needs through 2028.

This is, you know, the defining feature of the entire AI boom. The whole ecosystem is a web of massive, interconnected bets. Oracle's entire future is now tied to the business success of its tenants. And its main tenant, OpenAI, is a very hot, but very cash-intensive startup that needs to grow into one of the biggest companies on earth just to pay its rent.

And it's not just Oracle waiting on OpenAI to get rich. The scale of OpenAI's infrastructure ambition is staggering. It has a deal to buy tens of billions of dollars' worth of chips from AMD. It has an even bigger commitment to build out 10 gigawatts of Nvidia-powered systems—what Nvidia CEO Jensen Huang estimates to be four to five million GPUs.

Which raises the obvious, multi-hundred-billion-dollar question: how does OpenAI actually finance all of this? This exact question was put to Huang on CNBC recently, and his answer was surprisingly blunt:

They don't have the money yet. They're going to have to raise that money through, first of all, their revenues, which is growing exponentially, equity or debt.

This is, of course, the entire bet. Oracle's $185 billion forecast is built on the assumption that OpenAI can, in fact, grow exponentially and successfully raise all that equity and debt. The contract is a claim on a future that needs to be funded.

Oracle has survived a few cycles and has always emerged stronger. It is a survivor. But it has also never been a developer that burns cash for four years straight on a speculative building project. Which is fine, as long as the tenants' businesses keep booming and they can make their rent payments. But if the AI economy cools, Oracle could be left with the most expensive and most empty real estate in the world.

And this gets to the heart of why Oracle's story is so fascinating. The AI boom has officially stopped being a tech story and has become an industrial story.

The winners will no longer be determined just by who has the smartest algorithm, but by who can get the best terms on a multi-billion-dollar bond issuance and secure a priority connection to the power grid. Oracle's stock dipped not because its AI vision is weak, but because the market is starting to realize that the race to build AI infrastructure might be won not by the best engineers, but by the best plumber.

If you're interested in the second-order effects of this Oracle story—like the financial "plumbers" enabling these builds and the power companies racing to keep up—we have expanded on them in an extended analysis for our premium members.

Extended Analysis: The Plumbers and the Power Company

On Our Radar

Key signals we are tracking on ARPU Intelligence Desk

The Sanction-Fueled Champion

• The Headline: Chinese AI chipmaker Cambricon reported a 14x surge in quarterly revenue and swung to profitability as US sanctions blocking Nvidia's advanced chips have created a protected domestic market for local alternatives. (Bloomberg)
• ARPU's Take: The US ban on Nvidia's advanced chips has acted as a massive subsidy for its Chinese rivals. For Cambricon, the sanctions have created a captive, multi-billion dollar domestic market, essentially force-feeding it capital and customers to scale its technology. For Nvidia, this isn't just about losing the Chinese market; it's about the US government unintentionally incubating a formidable future competitor.
• The Implication: This is the first hard financial proof of the paradoxical effect of tech sanctions. While intended to handicap China's AI development, the policy is simultaneously creating a protected, non-competitive "greenhouse" for domestic champions to flourish. It signals a permanent bifurcation of the global AI hardware market and proves that geopolitical restrictions can be a powerful, if unintentional, form of industrial policy.

Meta's Off-Balance-Sheet AI Playbook

• The Headline: Meta has secured a record-breaking $30 billion financing deal by selling an 80% stake in its massive Hyperion AI data center to Blue Owl Capital, using a structure that keeps the immense debt off its own balance sheet. (Bloomberg)
• ARPU's Take: The capex for the AI arms race is so colossal that even Meta can't stomach it on its balance sheet. This deal allows them to secure the critical AI infrastructure they need to compete while pushing the enormous cost and debt onto a financial partner. For Blue Owl, it’s a landmark deal that turns a data center into a hot new investable asset class.
• The Implication: This deal creates the financial blueprint for the multi-trillion-dollar AI infrastructure buildout. It signals that the AI arms race is now a financial engineering race, not just a technological one. Expect every hyperscaler—Google, Amazon, Microsoft—to explore similar off-balance-sheet structures, turning the massive cost of data centers into a new product for Wall Street and private capital to buy.

These insights are drawn from our ongoing tracking at the ARPU Intelligence Desk. If you'd like a deeper dive into the full platform—including continuous signal monitoring and comparative analyses tailored for professional strategy work—we invite you to request a brief demo.