What is Oracle's AI data center backlog? Oracle's AI data center backlog is the pool of contracted, not yet recognized cloud revenue, reported as remaining performance obligations (RPO), that the company has signed with AI customers to deliver computing capacity over the coming years. On June 10, 2026, Oracle reported that this backlog reached a record $638 billion, up 363% year over year, the clearest single signal yet of how much physical data center capacity the AI boom will demand. For commercial real estate professionals, that number is not just an enterprise software headline; it is a forward order book for powered land, build-to-suit shells, and the megawatts that now drive site selection. For a broader view of how these tools fit together, see our guide to AI commercial real estate.
Key Takeaways
- Oracle's remaining performance obligations (RPO) backlog hit a record $638 billion in Q4 fiscal 2026, up 363% year over year and up $85 billion from the prior quarter.
- Oracle Cloud Infrastructure (OCI) revenue grew 93% to $5.8 billion, and the company plans to bring nearly one gigawatt of new computing capacity online in the current quarter alone.
- Analysts estimate more than half of the backlog traces to a single customer, OpenAI, making tenant concentration the central risk for anyone underwriting data center real estate.
- JLL projects global data center capacity will nearly double to roughly 200 gigawatts by 2030, with vacancy near 1% and most space precommitted before it is even built.
- For CRE investors, the lesson is to underwrite megawatts and tenant credit, not just square footage, and to stress test every deal against an AI demand pullback.
Oracle's AI Data Center Backlog Explained
Oracle's fiscal fourth quarter, reported after the market close on June 10, 2026, was a record across the board. Total revenue rose 21% to $19.2 billion, total cloud revenue climbed 47% to $9.9 billion, and OCI, the unit that rents out GPU capacity, grew 93% year over year to $5.8 billion. Non-GAAP earnings came in at $2.11 per share, up 24%.
The headline, though, was the backlog. RPO, the value of signed contracts Oracle has not yet delivered, jumped to $638 billion, a 363% increase over the prior year and an $85 billion gain in a single quarter. Oracle said nearly all of that incremental growth came from large-scale AI training contracts. CEO Clay Magouyrk told analysts the company expects to bring online almost one gigawatt of computing power in the current quarter, roughly its entire prior fiscal year. In real estate terms, one gigawatt can anchor several hyperscale campuses spanning hundreds of acres of powered land. Data center construction recently surpassed office construction spending in the United States for the first time, a shift we covered in our analysis of data centers overtaking office construction.
Why the OpenAI Concentration Is the Real Story for CRE Investors
Behind the record numbers sits a concentration risk every data center landlord should note. Bank of America analysts estimated that more than 50% of Oracle's backlog comes from a single customer: OpenAI, the maker of ChatGPT. Morningstar separately flagged customer concentration as a real risk given Oracle's outsize exposure to one counterparty.
For CRE investors, this is a familiar problem in unfamiliar clothing. A single-tenant data center anchored by one AI lab is, in underwriting terms, a credit play. The value of the lease depends entirely on the tenant's ability to pay over the full term, the way a net lease investor scrutinizes a drugstore or distribution tenant's credit. OpenAI only confidentially filed for an IPO this week and continues to burn cash even as revenue grows, a story we broke down in our coverage of the OpenAI confidential S-1 filing. If demand from a handful of frontier AI labs slows, the powered land and shells built to serve them could face re-tenanting risk far harder to solve than a vacant office floor.
What Oracle's Buildout Means for Data Center Real Estate
Oracle is not building in a vacuum. According to JLL's 2026 Global Data Center Outlook, global capacity is set to nearly double from roughly 103 gigawatts to 200 gigawatts by 2030, a buildout the firm values at more than $1 trillion in new real estate asset creation. Vacancy sits near 1%, and about 92% of capacity under construction is already precommitted through binding leases or owner-occupied development. As JLL's Andrew Batson put it, "Power has become the new real estate."
That reframes how CRE professionals value these assets. CBRE's 2026 U.S. Real Estate Market Outlook notes that developers now prefer 250 megawatt or larger projects on a single parcel with power delivery by 2028, and that grid interconnection timelines routinely stretch 24 to 48 months. Income is increasingly normalized around power, expressed as dollars per kilowatt per month, rather than dollars per square foot. The classic cap rate, net operating income divided by value, still applies, but the NOI now depends on how many megawatts a site can energize and how quickly. Powered land with a firm interconnection date has become one of the most valuable inputs in commercial real estate, a dynamic we explore in our guide to the AI data center power crisis and site selection.
How CRE Investors Should Underwrite Data Center Exposure Now
Oracle's backlog confirms the demand is real, but the concentration confirms the risk is too. Here is how disciplined CRE investors are responding in 2026:
- Underwrite the tenant, not just the building. For single-tenant AI deals, treat the lease like net lease credit. Model the tenant's funding runway, contract term, and what happens to the asset if the anchor walks.
- Price in interconnection risk. A site with power promised in 2028 carries execution risk that a stabilized asset does not. Confirm utility commitments in writing and tie rent commencement to actual energization.
- Favor diversified or hyperscaler-anchored deals. Build-to-suit projects backed by investment grade hyperscalers carry very different credit than speculative capacity leased to a single frontier lab.
- Stress test for a demand pullback. Run a scenario where AI capex growth slows and re-leasing takes 18 to 24 months. If the deal only works at full occupancy forever, it is not truly underwritten.
AI tools are now central to this work. Investors use models like ChatGPT, Claude, and Gemini to scenario test dollars per kilowatt economics, abstract power and lease agreements, and compare interconnection queues across markets. If you are ready to build that into your underwriting process, The AI Consulting Network specializes in exactly this kind of implementation.
The Bubble Question: Is the Demand Durable?
The market's reaction to Oracle's results was telling. Despite the record backlog, Oracle shares fell more than 7% in after-hours trading on June 10, as investors focused on the cost of delivering it. Oracle posted negative free cash flow of $23.7 billion for fiscal 2026, raised $43 billion in debt and $5 billion in equity during the year, and guided to roughly $70 billion in net cash capital spending in fiscal 2027, with reported capex higher still. The prepaid and customer-supplied GPU model, which now totals about $75 billion, softens that burden but underscores how capital intensive this race has become.
This is the same tension that prompted BlackRock CEO Larry Fink and Norway's sovereign wealth fund to warn this year that the AI investment cycle could produce bankruptcies and sharp valuation corrections. The lesson for CRE investors is not to avoid data centers; it is to respect that a $638 billion order book and a 7% stock drop can coexist. The infrastructure is being built faster than the revenue to justify it. CRE investors looking for hands-on AI implementation support, including durability stress tests for data center exposure, can reach out to Avi Hacker, J.D. at The AI Consulting Network.
Frequently Asked Questions
Q: What is Oracle's $638 billion backlog?
A: It is Oracle's remaining performance obligations (RPO), the total value of signed cloud contracts it has not yet delivered. Reported on June 10, 2026, it rose 363% year over year, driven almost entirely by large-scale AI computing contracts.
Q: Why does Oracle's data center buildout matter for commercial real estate?
A: Oracle plans to energize nearly one gigawatt of new computing capacity in a single quarter. That demand flows directly into powered land, build-to-suit data center development, and utility interconnection capacity, making it one of the strongest demand signals in commercial real estate today.
Q: What is the biggest risk in Oracle's backlog for CRE investors?
A: Tenant concentration. Analysts estimate more than half of the backlog comes from one customer, OpenAI. A data center leased to a single AI lab is effectively a credit investment, so the durability of that tenant's demand and funding is the key underwriting question.
Q: How should investors underwrite data center deals in 2026?
A: Underwrite megawatts and tenant credit, not just square footage. Confirm power delivery dates in writing, treat single-tenant AI deals like net lease credit, favor hyperscaler-anchored build-to-suit projects, and stress test every deal against a slowdown in AI demand.