Oracle's Backlog Tripled, Its Stock Fell Half. Here's What the Gap Means

Oracle (ORCL) holds $638 billion in contracted, not-yet-delivered AI cloud revenue. That number is roughly 1.4 times the company's entire $465 billion market value. And yet the stock is down about half from its 52-week high of $329.50, sitting near $161 as of this week.

That gap — a business booking record demand at a stock price that keeps getting punished — is the whole fight. The report comes out Thursday, September 10, after the close, and the options market is pricing a move of roughly 11%, calmer than the stock's own average swing of about 16% over the last four quarters. The calm itself is a clue: investors are no longer betting this earnings call surprises them on demand. They're betting on something else entirely.

The numbers on the table

Here is the shared record both bulls and bears can sign, all from Oracle's own disclosures and this week's consensus:

• The backlog. Oracle's "remaining performance obligations" — contracts signed but not yet delivered — hit $638 billion last quarter, up 363% over a year and up $85 billion in a single quarter.
• The growth. Oracle Cloud Infrastructure (the AI compute cloud) grew 93% in the last reported quarter to $5.8 billion. Analysts expect revenue of about $19.1 billion this quarter, up roughly 28% year over year.
• The burn. Last full year Oracle generated a record $32 billion of operating cash flow, but its free cash flow — the cash left after paying for the data centers — was negative $23.7 billion, because it spent $55.7 billion on the buildout.
• The next bill. Management has guided to $70 billion to $95 billion of capital spending this year, and said it expects to raise roughly another $40 billion, in a mix of new debt and new stock.

Two concepts do the heavy lifting here. Backlog (or RPO) is a promise of future revenue, not cash in the bank today. Free cash flow is what's left after a company pays for everything it needs to grow — for Oracle right now, that's the expensive data centers that will eventually deliver on that backlog. Keep those two straight and the entire debate snaps into focus.

Round 1 — The bull's best punch

The bull's argument is a simple one about visibility. When a company signs $638 billion in contracts, the demand is not a forecast — it is a signature. The cloud infrastructure business growing 93% is not a low-single-digit recovery; it is the fastest engine in the company, and it is the part of Oracle that the market pays up for.

Add in a real offset to the spending: $75 billion of that backlog already arrived as customer prepayments or customer-supplied hardware, which means some buyers are paying for the data centers up front rather than leaving Oracle to fund them. And then there is the price. After a 50% slide, the stock is at a level a handful of analysts called "an attractive entry," with 36 buy ratings against a single sell and an average price target near $242 — more than 50% above where it trades.

The bull wins this round on a clean fact: the backlog-to-market-cap ratio (about 1.4 to 1) is a genuine measure of cheapness the bears cannot easily explain away. A company with $638 billion of contracted future revenue worth less than half its market value is not priced like a company in trouble.

But there is one number that genuinely embarrasses the bull case, and the bear is waiting for it: management says only about 12% of that backlog converts to revenue in the next 12 months. Most of the $638 billion is a two- to three-year promise.

Round 2 — The bear's best punch

The bear doesn't need to argue that demand is fake. The bear argues that a promise is a liability when it's concentrated and it's expensive to keep.

The concentration is the sharpest point. Roughly half of the $638 billion backlog is attributed to a single customer — OpenAI. OpenAI is not a profitable, cash-rich anchor tenant. It reported a $20.9 billion operating loss last year, and its long-rumored IPO is reportedly pushed to 2027. So the single biggest driver of Oracle's growth story is, in the bear's telling, a customer that may need to stay funded for years. One customer, one unprofitable balance sheet, one IPO that keeps slipping — that is a single point of failure dressed as a moat.

Then the financing. The AI buildout is cash-negative now: negative $23.7 billion of free cash flow last year, and rating agency S&P projects a wider deficit of roughly $42 billion this year. To close the gap, Oracle is raising about $40 billion — including a $20 billion drip of new stock (an "at-the-market" issuance) plus a $5 billion mandatory convertible preferred. That is real dilution layered on a heavy debt load. The June quarter is the perfect exhibit: Oracle reported a record $19.2 billion in revenue, and the stock more than 8%. The market had already heard the growth; it was reacting to the spending.

The bear wins this round on mechanism: growth that must be funded by new debt and new shares, and that depends on one customer's ability to pay, is a different animal than growth that pays for itself.

The bear's honest concession: this is not a startup with a questionable business model. Oracle is a 50-year-old franchise with $32 billion of annual operating cash flow, a real installed base, and — in the bull's own best form — 72 data centers embedded inside Microsoft's and Amazon's clouds, collecting revenue no matter which big player wins a given workload.

What the price is actually betting on

Now make both stories pay rent, and the fight reduces to a single variable: conversion.

Can Oracle turn that $638 billion into cash faster than the debt and dilution accumulate? That is not a question this quarter's revenue beat can answer, because only 12% of the backlog shows up in the next year. The number to watch is not the headline; it is the trajectory of free cash flow and the share of new contracts coming from OpenAI.

And that is why the options market is only asking for 11%. A year ago, the fear was that the AI boom might stall and demand could disappoint — that would have warranted a bigger two-way bet. The demand has, in fact, been extraordinary. So the market has moved on. The residual 11% is a price on the funding and concentration risk, not on whether customers want the product. The bet is no longer "will the backlog materialize?" It is "will Oracle stay solvent and keep control of its own shares while it does?"

That reframes the valuation. A stock down 50% that analysts rate at 50% upside can still be a high-conviction leveraged bet on one thing going right. The question for the buyer is not whether Oracle is a great company — the evidence says it is growing fast — but whether the price has already charged enough for the two risks that could break the machine.

The ruling

Bull on the business. Careful on the stock at this price.

The business case is genuinely strong: fastest-growing major cloud, $638 billion of contracted demand, and a valuation that looks cheap against that backlog. That is a real, defensible long, and the 50% drawdown has removed a lot of the froth from June.

But the stock, at this level, is still a leveraged bet on OpenAI staying solvent and on Oracle converting a mostly two- to three-year backlog into cash ahead of a $70–95 billion spending bill. A 50% fall is context, not a margin of safety — it priced in the debt fear, but it did not remove the two live risks, and the growth still has to outpace the burn for the cheapness to prove out. The burden of proof sits with the bulls, and it is cash collection, not another contracted-revenue headline.

The reversal clause. Watch two quarters from here — through the December report and the March 2027 report. If Oracle Cloud Infrastructure growth stays in the guided 58% to 64% band and free cash flow stops bleeding, the bull's "funded growth" story holds and the debt narrative fades. The ruling flips the other way if OpenAI keeps absorbing an ever-larger share of new backlog while free cash flow runs deeply negative into early 2027: that combination is the bear's single point of failure becoming a balance-sheet problem, and it would mean the cheapness was a trap, not an entry.