Oracle says $30 billion. Analysts said $6 billion. We ran the math

Short answer: Oracle’s $30 billion can come true, but only if Oracle cuts its building spending by half or more within two years. Today Oracle spends more on data centers than it makes. On October 28 it promised that by fiscal 2029 (June 2028 to May 2029) it will have $30 billion of real cash left over every year. Analysts had expected about $6 billion. The difference is not about how much Oracle sells. It is about how fast it can stop building. And part of Oracle’s cash is money customers pay in advance, which makes the number look better than the business behind it.

Oracle's free cash flow in fiscal 2029 (June 2028 to May 2029)

Oracle says$30B+target set at Investor Day, Oct 28

vs

Analysts expected$5.9Bconsensus before that day (TIKR)

The $24.1 billion gap is less than what Oracle spends on data centers in one quarter today.

Where it is now

−$23.7Bfree cash flow in fiscal 2026 (to May 2026)

Spent on data centers

$55.7B

First quarter of fiscal 2027

−$5.4B

What $30B needs

$29–49Bthe most Oracle could spend on building in fiscal 2029

Planned this year

$90–95B

Cut needed

47–69%

While sales keep growing

$166Bfiscal 2029 revenue on the way to Oracle's $225B goal for fiscal 2030

This year

$90B+

Growth a year

36%

The catch

$11.4Bpaid in advance by customers last quarter, counted as cash Oracle made

Reported free cash flow

−$5.4B

Without the prepayments

−$16.8B

1. One company, two numbers

Sources: Oracle annual results via Yahoo Finance (fiscal 2023–2026); Oracle Investor Day release, Oct 28, 2026 (SEC 8-K); TIKR consensus, Oct 29, 2026.

Free cash flow is the simplest test of a business: the cash it brings in, minus what it spends on buildings and equipment. It is what is left to pay debts, pay shareholders, or save. In fiscal 2023 and 2024, Oracle had $8–12 billion of it a year. Then the AI boom started. Oracle signed huge contracts to rent computing power to AI companies, and to deliver them it has to build data centers first. In fiscal 2026 (the year to May 2026) it spent $55.7 billion on building and ended the year $23.7 billion short.

Investors read this in two ways. Stock investors see a company that has already signed $664 billion of future sales and trades at only about $10 for every dollar of next year’s profit, less than Microsoft, Amazon or Google. Bond investors, who lend Oracle money, see a company borrowing to build: insuring Oracle’s debt against default now costs almost four times more than a year ago, and S&P rates it one step above “junk”.

On October 28, at its Investor Day, Oracle answered with a number: free cash flow turns positive in the second half of fiscal 2028 and reaches at least $30 billion in fiscal 2029. The stock jumped 11% that day. Analysts tracked by TIKR had expected $5.9 billion for that year. Some reports say analysts expected free cash flow to still be negative. Either way, Oracle’s number is far above what Wall Street had.

2. The math: the building has to slow down, a lot

Sources: Oracle annual results via Yahoo Finance; fiscal 2027 plan from the Sep 10, 2026 earnings call (CNBC, erp.today); fiscal 2029 ranges are our estimate, see Methodology.

Free cash flow has two parts: cash in and cash out. So we checked what each number needs.

Cash in. Oracle says it will sell at least $90 billion this fiscal year and $225 billion or more in fiscal 2030. Growing evenly between the two, that is about $122.1 billion in fiscal 2028 and $165.8 billion in fiscal 2029: 36% more every year. Not all of that becomes cash. In fiscal 2023–2025, Oracle’s cash from operations was about 35% of its sales. In fiscal 2026 it was 47.5%, partly because customers started paying in advance. That gives fiscal 2029 cash from operations of about $59–79 billion.

Cash out. Take $30 billion away and what is left is the most Oracle can spend on building: $29–49 billion. This year Oracle plans to spend $90–95 billion. So spending has to fall by 47–69% in two years.

The analysts’ $5.9 billion is the same math with more building: about $53–73 billion of spending. That is the whole duel. Stocks bet that the building slows down fast. Bonds worry it does not.

Here is why that is hard. This year Oracle spends about $1 on building for every $1 of sales. For $30 billion it needs 17–29 cents, in the same years its sales have to keep growing. To reach $225 billion in fiscal 2030, sales still have to grow by about $59 billion the year after fiscal 2029, and new sales need new data centers. Before the AI boom, in fiscal 2024, Oracle spent 13 cents per dollar of sales.

3. Why Oracle thinks it can do it

Oracle has a real argument. A data center costs most while it is being built, and earns for years after. The ones Oracle is building in fiscal 2027 and 2028 start bringing in rent later, without the building cost. Oracle says spending peaks in fiscal 2027–2028 and steps down in fiscal 2029. It did not say by how much.

It also plans to make other people pay part of the bill. In Oracle’s words: “Vendor financing, customer-provided hardware and customer prepayments are expected to reduce upfront cash requirements.” In plain words: suppliers wait to be paid, some customers bring their own chips, and some pay in advance. Some analysts believe the plan. In March, a Guggenheim analyst said Oracle’s cash “turns into a free cash flow waterfall in fiscal 29 and 30”.

4. The catch: part of the cash is customers’ money in advance

Source: Oracle results releases for fiscal 2026 (Jun 10, 2026) and Q1 fiscal 2027 (Sep 10, 2026), SEC 8-K; 'without' is our arithmetic.

When a customer pays upfront for computing it will use later, Oracle counts that money as cash from operations, so it raises free cash flow. In the last quarter alone, customers prepaid $11.4 billion. With it, Oracle’s free cash flow was −$5.4 billion. Without it, it would have been −$16.8 billion.

The money is real: it sits in Oracle’s bank account. But it is not free. Oracle owes those customers years of computing in return, and it still has to build the data centers to deliver it. This is where stocks and bonds disagree again. Shareholders see cash. Lenders see a promise Oracle still has to keep. If a big part of the $30 billion in fiscal 2029 comes from prepayments, the number can be hit while the business itself is still spending more than it earns.

5. Run the math yourself

Revenue

$166B

Oracle's path: about $166B. Fiscal 2026: $67.4B.Cash share of revenue

47.5%

Fiscal 2023–2025: about 35%. Fiscal 2026: 47.5% (with prepayments).Spent on building

$92.5B

Plan for this year: $90–95B. Fiscal 2024: $6.9B.

Free cash flow

$6B$30B

Move the sliders. Two things stand out. First, sales matter less than you would think: even at $225 billion of sales and last year’s cash share, Oracle cannot make $30 billion while it keeps building at this year’s pace. Second, the cash share matters a lot. Five points of it on $166 billion of sales is about $8 billion, a third of the gap between the two numbers. That is where prepayments come in.

6. What to watch

• Mid-December 2026, second-quarter results. Is spending still on track for $90–95 billion this year, and how much of the cash came from prepayments?
• June 2027, fiscal 2028 outlook. How high is the peak? A higher peak means a steeper drop is needed later.
• December 2027 to May 2028. Oracle says free cash flow turns positive in these months. This is the first hard test.
• June 2028, fiscal 2029 outlook. The step-down in spending gets a number. If it is above about $49 billion, $30 billion becomes very hard.

Oracle ORCL

At publication

$116.10

Nov 6, 2026 close

→

Now

…

Loading live price

Difference since publication

…

Source: Yahoo Finance daily closes. Live data refreshes on page load.

The price was frozen at the close on Friday, November 6, at $116.10. The live price loads every time the page opens. The stock jumped from $130.53 to $144.54 on the day of the promise, and has fallen about 20% since.

My take: neither number is a fantasy. The $30 billion is not about selling more; Oracle already has the contracts. It is a bet that Oracle can almost stop building new data centers just one year after its biggest building spree, and that customers keep paying in advance. Oracle has not said how much it will spend in fiscal 2029, and that is the number that decides the duel. Until it does, the $30 billion is a promise, and the bond market prices it like one. Which number do you believe?

What the numbers leave out

• Our revenue path is a straight line. Oracle gave only fiscal 2027 and fiscal 2030. If sales grow faster early, fiscal 2029 cash is higher; if they grow later, lower.
• The cash share can change. As data centers age, depreciation (a cost on paper, not in cash) grows, which can lift cash from operations above its old share. Higher interest on Oracle’s debt pushes the other way.
• Not all spending is the same. Oracle also reports “net” building spending, after money from customers and short-term financing: $47.7 billion in fiscal 2026, against $55.7 billion in total. For this year, it guides net spending to no more than $70 billion. Free cash flow uses the total.
• Customers and politics. About half of Oracle’s contracts reportedly come from OpenAI, which still loses money. And after the November 3 elections, Texas, home of Oracle’s biggest AI campus, elected a governor who wants to pause new data centers.
• The funding cap. On October 28, Oracle promised to raise no more than $20 billion of new money through fiscal 2030. On November 2, it filed to sell $10 billion of shares and $10 billion of bonds. If that counts toward the cap, Oracle has little room left if the cash comes later than planned.
• The consensus. $5.9 billion is TIKR’s consensus of the analysts it tracks. Other reports say analysts expected free cash flow to still be negative in fiscal 2029.

Data

Download the yearly figures and both fiscal 2029 scenarios: oracle-cash-math.csv.

Methodology

• Oracle’s targets: Investor Day release of October 28, 2026, SEC 8-K Exhibit 99.1: free cash flow of “$30 billion+ for FY29”, positive in “the second half of FY28”, “$225 billion or more in FY30”, spending that is “expected to step-down in FY29”, the $20 billion funding cap and the quote on vendor financing, customer-provided hardware and prepayments. The analysis script checks every quoted phrase against the filing.
• Past results: yearly revenue, cash from operations and capital spending for fiscal 2023–2026 from Yahoo Finance (yfinance), matched with Oracle’s fiscal 2026 results release (SEC 8-K): cash from operations $31,977M, capital spending $55,663M, free cash flow −$23,686M, prepayments $4,592M, net building spending $47,726M.
• Last quarter: Q1 fiscal 2027 release (SEC 8-K): revenue $19,345M, cash from operations $23,103M including $11,363M of customer prepayments, capital spending $28,499M, free cash flow −$5,396M. “Without prepayments” is our arithmetic: reported free cash flow minus prepayments.
• This year’s plan: capital spending of $90–95 billion and net spending of no more than $70 billion, from the September 10, 2026 earnings call, reported by erp.today and CNBC.
• The model: fiscal 2029 revenue = $90B × (225 / 90)^(2/3) ≈ $165.8B (constant growth between Oracle’s two targets). Cash from operations = revenue × cash share; we use two shares from Oracle’s own history: 35.3% (average of fiscal 2023–2025) and 47.5% (fiscal 2026). Allowed spending = cash from operations − target. These are scenarios to show what each number needs, not a forecast.
• Consensus: $5.9 billion for fiscal 2029 from TIKR (Oct 29, 2026); “analysts had expected the figure to still be negative” from The Next Web. Guggenheim quote: 24/7 Wall St..
• Bonds and rating: default-insurance (CDS) prices from Securitas Global and CNBC (Oct 19, 2026); S&P rating BBB−. Price per dollar of profit: forward P/E from Yahoo Finance, November 6, 2026.
• Other events: the $20 billion share and bond sale filed on November 2 (CNBC); Texas election (CNBC); OpenAI’s share of contracts is a press estimate, not disclosed by Oracle.
• Share prices: daily closes of ORCL from Yahoo Finance via yfinance (unadjusted). Prices at publication were saved with yfinance; live prices come from a small server function that queries Yahoo Finance when the page loads.
• Code: plain Python scripts (collect.py, analyze.py, snapshot_stocks.py).
• Trademarks: company and product names are trademarks of their respective owners. This site is independent and not affiliated with or sponsored by Oracle, OpenAI or any other company mentioned.