Oracle’s backlog hit $664 billion. It sold $20 billion of stock to build it

Oracle opened Friday at $164.43, traded as high as $166.00, and closed at $150.28. Nothing happened in between except that investors had time to read the release.

The headline was extraordinary. Oracle’s remaining performance obligations — the contracted work it has not yet delivered — reached $664 billion, up $209 billion in a year. Cloud infrastructure revenue grew 121%. Total revenue rose 30% to $19.3 billion. Wire coverage led with a 7% surge, and for about an hour that was accurate.

By the closing bell the entire gap had gone, and Oracle finished the week down 5.4%. The stock that booked the largest backlog in the AI industry was one of the week’s most heavily traded names and one of its more instructive declines.

Article Brief

Key Takeaways

5 Points30s Read

1. 01The reversal–Oracle opened Friday at $164.43 and closed at $150.28 after touching $166.00, finishing the week down 5.4% on roughly 80 million shares.
2. 02The quarter–Q1 FY27 revenue rose 30% to $19.3 billion, cloud infrastructure grew 121% to $7.4 billion, and RPO reached $664 billion, up $209 billion year over year.
3. 03The line nobody led with–Oracle sold $20 billion of common stock through an at-the-market programme during the quarter, while free cash flow ran at negative $5 billion.
4. 04The scale–The backlog equals roughly 7.4 years of Oracle’s own full-year revenue guidance of at least $90 billion. RPO is not revenue.
5. 05The concentration–Roughly half the backlog is tied to OpenAI, per reporting on the quarter, though the non-OpenAI portion has more than doubled in a year.

The quarter was genuinely excellent

This is not a story about a bad quarter, and any version that pretends otherwise is wrong.

Oracle’s own release reports first-quarter GAAP earnings per share up 55% to $1.56 and non-GAAP up 30% to $1.92. Total cloud revenue rose 62% to $11.6 billion. The infrastructure line — the part that competes with Amazon, Microsoft and Google for AI workloads — grew 121% to $7.4 billion. GAAP operating income rose 57%.

The delivery numbers are equally real. Oracle says it brought 850 megawatts of additional datacenter capacity online in the quarter and delivered more than 300,000 GPUs to AI cloud customers, almost triple the capacity it delivered in the prior quarter. Full-year guidance now calls for at least $90 billion of revenue.

A company does not fake 850 megawatts. Whatever the market was reacting to, it was not doubt about execution.

The rest of the business behaved as it has for years. Software revenue fell 3% to $5.5 billion as customers keep migrating off on-premises licences, services rose 5% to $1.4 billion and hardware rose 15% to $0.8 billion. Oracle is now a company where one line item does all the growing, and that line item happens to be the most capital-hungry business it has ever operated.

Guidance did not wobble either. Oracle expects second-quarter revenue growth of 30% to 34% and cloud revenue growth of 64% to 70%, with full-year non-GAAP earnings per share of $8.10. On the numbers alone, the reaction on Friday makes no sense at all.

What was in the release but not in the headline

Two sentences in the same announcement explain the afternoon better than any analyst note.

The first: “Free cash flow was negative $5 billion for Q1 as Oracle continued to execute on investments to support the growth of its Cloud Infrastructure business.” That sits directly beneath a record operating cash flow of $23 billion, up 184%. Oracle generated enormous cash from operations and still spent several billion more than it produced.

The second is the one that got almost no coverage: “During Q1 FY 2027, Oracle successfully completed the sale of $20 billion of common stock (before commissions) through an At-the-Market (ATM) equity program, as part of its previously disclosed capital investment program.”

Oracle sold $20 billion of its own shares during the quarter.

An at-the-market program does not announce itself the way a follow-on offering does. It drips stock into the open market over time, at prevailing prices. It is an efficient way to raise money and a quiet one, and it means existing holders were being diluted throughout the period in which the backlog they are now celebrating was being built.

Read the release in order and the sequence is plain: record demand, negative free cash flow, twenty billion dollars of equity issued. The backlog is being financed, and shareholders are one of the financiers.

In fairness to Oracle, none of this was sprung on anyone. The company published an equity and debt financing plan for calendar 2026 back in February, and the release describes the stock sale as part of that “previously disclosed capital investment program.” It also went out of its way to state that the structuring of the new contracts means “there is no incremental impact on its plans to raise capital” — a pre-emptive answer to a question the company clearly expected. Analysts have started describing the result as a new financing model rather than simply a strong quarter, which is closer to the truth.

The distinction between disclosed and priced is where Friday lived. Investors knew a financing plan existed. Seeing $20 billion actually executed, in the same paragraph as a negative free cash flow number, is a different experience from knowing it was possible.

Seven years of revenue, already booked

The $664 billion figure is where the enthusiasm lives, so it is worth sizing properly.

Oracle guides to at least $90 billion of total revenue for the full fiscal year. Against that, the backlog represents roughly 7.4 years of the entire company’s revenue — not 7.4 years of cloud revenue, but of everything Oracle sells, including the software and hardware businesses that have nothing to do with AI.

That is not a criticism. Long-dated infrastructure contracts are supposed to look like this. But it does clarify what the number is and is not. RPO is not revenue, it is not cash, and it converts only as fast as Oracle can physically build capacity to serve it. Oracle said so itself: demand for AI training and inferencing “continues to grow faster than supply.”

When demand exceeds supply, the constraint is not the order book. It is megawatts, GPUs, construction timelines and the money to fund all three. The company added 850MW and 300,000 GPUs in a quarter, and still the backlog grew faster than the delivery.

One piece of arithmetic is worth doing slowly. Oracle ended its previous quarter with $638 billion of RPO and ended this one with $664 billion — a sequential increase of $26 billion. In the same quarter it says it booked more than $30 billion of new AI cloud contracts. The gap between those two numbers is roughly what got delivered and recognised as revenue along the way. That is the conversion engine working exactly as intended, and it is also a reminder of the scale involved: at that rate of burn-down, a backlog of this size is a very long-dated asset, and long-dated assets are the ones most exposed to a counterparty’s fortunes changing in the meantime.

Half of it rests on one customer

The concentration is the part that has drawn the most analyst attention, and it is not disclosed in the release.

Roughly half of the $664 billion backlog is tied to OpenAI, according to reporting on the quarter. Oracle’s non-OpenAI backlog has more than doubled over the past year, which genuinely does broaden the base — but it does not change the arithmetic that a single counterparty sits behind an enormous share of contracted future work.

TECHi has been circling this all week from other directions. We wrote about a hedge fund that held the right AI thesis with too much borrowed money and lost two-thirds of its value in a month, and about Anthropic’s proposed listing, where the anchor investor being courted is also the supplier being paid. Oracle is the third shape of the same structure: a supplier booking enormous forward revenue from AI labs that are themselves consuming capital rather than generating it, and funding the buildout with issued equity in the meantime.

None of these is fraud or even imprudence. They are what a capital-intensive boom looks like from three different seats.

The read-through was immediate. CoreWeave and Nebius each rose about 4% on Oracle’s numbers, because a $664 billion backlog at the largest player validates the demand thesis for every neocloud behind it. That sympathy trade is rational on demand and awkward on funding: the smaller operators cannot sell $20 billion of stock into a rally, which is precisely the constraint TECHi examined in CoreWeave’s debt-funded backlog.

Oracle closed Friday at $150.28 on roughly 80 million shares, its heaviest volume in months. That leaves it about 39% below where it closed on June 1, which matters for reading the reaction: this is not a crowded long giving back a little froth, it is a stock that has already spent a summer de-rating and could not hold a 7.5% gap on its best operational quarter in years.

TECHi flagged the shape of this test months ago, when we argued that the backlog was huge but AI capex was the harder test. The market has now priced the second half of that sentence.

What would change the read

The bullish resolution is straightforward and entirely plausible: Oracle keeps converting backlog into delivered capacity at the rate it just demonstrated, free cash flow turns as the build matures, and the equity issued in 2026 looks cheap against the revenue it financed. Triple-digit infrastructure growth buys a lot of patience.

The bearish one does not require the AI trade to break. It only requires conversion to run slower than the financing. Every quarter of negative free cash flow is another quarter of funding decisions, and the release’s careful note that new contracts carry “no incremental impact on its plans to raise capital” is the sentence of a company that knows the question is coming.

What to actually watch is narrow, and it is not the backlog. It is free cash flow, whether the ATM program is extended, and the pace of megawatts delivered per quarter. Those three tell you whether $664 billion is a schedule or an aspiration. Friday’s reversal suggests the market has started grading the same three.