Oracle vs Microsoft AI Cloud Spending: Cash Flow, Capex and Creative Financing

Oracle's latest quarter sits in the middle of that argument. Microsoft's own disclosures about data center accounting are relevant too, and many coverage pieces leave them out.

Here is what is verified, what is only attributed opinion, and why it matters if you run Azure, Microsoft 365 or Oracle Cloud workloads.

The Two Positions​

According to 24/7 Wall St., Odd Lots guest Luke Kawa argued that this boom is unusual because AI's own profits aren't paying for it. In his telling, Microsoft and the other top spenders have cash machines of their own, and Oracle doesn't have that "pure money spigot." Host Joe Weisenthal reportedly cited estimates that a few companies would need revenue equal to 9% of GDP to justify current spending. Kawa reportedly added that about half of this year's expected earnings growth comes from hyperscaler capital spending.

Treat those as claims made on a podcast, not as established facts. I could not independently check the quotes, the 9% figure or the earnings-growth estimate against a transcript.

Cramer's side, also as reported by 24/7 Wall St.: Microsoft chose to accept some short-term strain on its balance sheet because data center assets last longer than skeptics think. He also pointed to Copilot adoption and faster Azure growth. One correction: Cramer reportedly said Copilot has 30 million "users," but Microsoft's figure is more specific. Satya Nadella's FY2026 Q4 statement says Microsoft 365 Copilot reached more than 30 million paid seats.

Section summary: One side says AI spending is being paid for with outside money. The other says the returns are starting to arrive. Company filings are the best way to test both.

Oracle's Quarter: Growing Fast and Burning Cash​

Oracle's fiscal Q1 2027 results, for the quarter ended August 31, 2026, explain why it gets singled out.

The growth numbers are very strong:

• Total revenue was $19.345B versus $14.926B a year earlier, up 30%; total cloud revenue was $11.607B, up 62%; and cloud infrastructure (IaaS) reached $7.388B, up 121%.
• Remaining performance obligations (RPO) hit $664 billion, up $209 billion year over year and $26 billion quarter over quarter, versus StreetAccount consensus of $630.6 billion.
• Oracle booked more than $30 billion of additional AI cloud contracts in Q1.
• Management said data center delivery is speeding up, with 850 megawatts of capacity added during the quarter and AI infrastructure utilization near 98%.

The cash side is the problem:

• Operating cash flow was $23.103 billion (+184%), driven by earnings growth and customer prepayments. Capex was $28.499 billion versus $8.502 billion a year ago, leaving free cash flow at −$5.396 billion.
• A year earlier, the deficit was only negative $362 million.
• Oracle beat on revenue, EPS and backlog and raised full-year guidance, and the stock still fell 5.4% during the 10 September session.

The funding plan, then. Oracle completed a $20 billion at-the-market common stock sale during the quarter. 24/7 Wall St. says it plans to raise about $40 billion in debt and equity this fiscal year, and that interest expense rose 55% to $1.4 billion. That $40 billion is a mix of debt and equity, not all borrowing, and many retellings blur that point.

Section summary: Oracle has the fastest infrastructure growth among the big clouds, a huge backlog and nearly full utilization. It is also spending more than its record operating cash flow and selling shares to cover the difference.

How Oracle Is Funding It​

Oracle is changing how it pays for capacity, not just how much it spends. CFO Hilary Maxson said the vast majority of the more than $30 billion in new AI contracts signed during the quarter came through prepayment, bring-your-own-hardware, or similar structures, so they will not require incremental capital from Oracle. ERP Today reports that CEO Clay Magouyrk went further, arguing that Oracle's own capex should now be considered decoupled from how fast the business can grow, with suppliers and customers sharing the capital load. That fits the line 24/7 Wall St. attributes to him: "It doesn't have to be Oracle CapEx."

Oracle now reports a figure to show this. It publishes a newer non-GAAP measure, "net cash outlay for capital expenditures", which subtracts those prepayments from capex and came to $18.0 billion for the quarter. For the full year, management still expects $90 billion to $95 billion of total CapEx in fiscal 2027, capped at $70 billion on a net-cash basis after alternative financing.

My view, as an industry observer: customer prepayments are money Oracle owes back in services, not profit. The cash comes in now, but Oracle still has to deliver the capacity later. FinanceFeeds, reading Oracle's 10-Q, puts it this way: when those contracts convert, Oracle books the revenue and margin, but the cash arrived in fiscal 2027's first quarter and was absorbed by data center spending. The same filing shows Oracle expects about 50% of that backlog to become revenue within 36 months, up from 35% a year earlier. That counts in favor of Oracle's case. Still, the filings do not split the $664 billion by payment structure.

And there's no timetable for getting back to positive cash flow. Oracle declined to say when cash flow will turn positive, leaving the Q2 FY27 report and the October investor day as the next tests.

Microsoft: Positive Cash Flow, and an Accounting Change​

Microsoft is different, though not as different as the "safe vs. risky" framing implies.

Results from Microsoft's FY2026 Q4 release (quarter ended June 30, 2026):

On the call, CFO Amy Hood said about two-thirds of quarterly capex went to short-lived assets, mainly CPUs and GPUs. Finance leases totaled $5.6 billion, mostly for large data center sites. Nadella said Azure passed $100 billion in annual revenue for the first time. 24/7 Wall St. also reports full-year capex of $115.95 billion, full-year free cash flow down 6.46% to $66.99 billion, and Hood's expectation that Microsoft will "remain free cash flow positive in FY27." I couldn't check those full-year figures or that quote against the primary materials, so they rest on 24/7 Wall St.'s reporting.

The more important detail for the "creative financing" debate is in Hood's prepared remarks:

1. Longer asset lives. Starting in FY27, Microsoft is extending the estimated useful lives of its data centers and office buildings from 15 to 25 years. Hood said this changes only the timing of depreciation and is expected to give a minimal benefit to FY27 operating income.
2. A shift from finance leases to operating leases. Hood said the bigger effect is on reported capex, because more future data center leases will become operating leases instead of finance leases. Finance leases count toward capex. Operating leases do not.
3. A lower headline capex number. Apart from the useful-life change, Microsoft's calendar 2026 capex plans are unchanged. The lease reclassification moves the expectation to about $175 billion. Microsoft guided FY27 Q1 capex to more than $50 billion, including the reclassification.

None of this reduces the money Microsoft spends building data centers. What changes is which line of the financial statements the spending appears on, and when depreciation hits earnings. So when the reported capex number changes, check whether the lease classification moved before deciding spending actually changed. 24/7 Wall St. is right that this is where the financing debate applies to Microsoft too.

Section summary: Microsoft funds its buildout from strong operating cash flow and still produces free cash flow. Its accounting changes will make the reported capex figure look different, though, and that is worth watching.

Why IT Pros Should Care​

This is more than a stock story. The financing pressure connects to things admins deal with every day. This part is my own analysis:

• Capacity is still the bottleneck. Hood said Azure customer demand "continues to exceed available capacity," and Oracle reports near-full AI utilization. If you're planning GPU workloads, expect reserved capacity, commitments and regional availability to stay tight.
• Pricing is moving toward usage-based billing. On the same call, Microsoft said GitHub Copilot moved to usage-based pricing in June, and Microsoft 365 added usage-based billing products alongside per-seat licensing in July. When infrastructure costs this much, vendors want revenue that scales with consumption. Budget for less predictable bills.
• Concentration risk for OCI customers. StorageNewsletter's commentary notes that Oracle's customer base remains concentrated on a few AI labs, OpenAI above all. If you're an enterprise OCI customer, it's fair to ask how dependent your vendor's roadmap is on a few huge tenants.
• RPO is not cash. Microsoft says its RPO has a weighted average duration of 2.3 years, with about 30% recognized in the next 12 months. A backlog is a promise of future revenue, not money the company already has.

What Would Settle the Argument​

24/7 Wall St. names three signals, and they hold up:

1. Which way free cash flow goes. Does Oracle's deficit shrink without new equity or debt, and does Microsoft's free cash flow stay positive while capex rises?
2. How the money is raised. How much of Oracle's roughly $40 billion plan ends up as debt, and how much of Microsoft's buildout moves into operating leases?
3. Whether revenue speeds up. Microsoft guided Azure growth of about 45% in constant currency. Oracle guided Q2 cloud growth of 65% to 71% in U.S. dollars. Revenue growing faster than spending would close the gap fastest.

The Bottom Line​

Neither side has won yet. Oracle's numbers support both views: real demand for its infrastructure and a cash shortfall covered with equity, customer prepayments and supplier arrangements. Microsoft's numbers show real cash generation alongside accounting changes that will make its reported capex look smaller. Share prices can't tell you whether a funding structure is wise, and 24/7 Wall St. itself concedes the price gap proves nothing about cause.

The useful distinction is this: Microsoft funds its AI buildout mostly from operating cash flow, while Oracle has to rely more on outside money. That's a description of where each company stands today, not a forecast of which one wins.