For most of this year, the market has treated Oracle's $638 billion backlog as the answer to every doubt about AI spending. Big number, case closed, next question.

Then you look at who is actually behind that number.

Roughly half of Oracle's backlog traces back to a single company. And that company spent part of this year missing its own internal targets.

That is not a footnote. That is the entire bull case standing on one leg.

Today I'm cross-examining Oracle's $638 billion promise, because a backlog is only as good as the customer paying it off, and I don't think enough people have actually asked who that customer is.

Missed Monday's edition on Palantir? 😲 The Stock That Outran Its Own Analysts

The big takeaway from Jackson Hole is that Fed Chair Kevin Warsh sounded more hawkish than markets expected. Analysts now see a higher chance of a September rate hike, with Reuters reporting markets were pricing more than a 65% probability of a 25-basis-point move after the speech. In plain English, Warsh is trying to reassert inflation-fighting credibility, and that means investors may need to take “higher for longer” seriously again.

CrowdStrike’s rally is being driven by more than just momentum. The company sits right in the middle of a spending priority that is getting harder to ignore: securing AI-era endpoints, identities, cloud workloads, and automated threats. The caution is valuation, because the stock is now trading at a steep premium, but the bull case remains that Falcon is becoming a core security platform at exactly the time enterprises are being forced to upgrade cyber defenses.

Both dollar-store stocks look better, but the drivers are not identical. Dollar General is benefiting from budget-stretched shoppers trading down, with same-store sales up 3.5%, while Dollar Tree is showing early signs of traffic recovery after selling Family Dollar and simplifying the business. The bigger message is that the lower-income consumer remains under pressure, but that pressure can actually become a tailwind for the best-positioned value retailers.

"When I think of revenue, I think of Attio." - Shreman Shrestha, Head of Business at Granola

Here's what that adds up to:

Start Your Trial

ORCL's remaining performance obligations hit $638 billion last quarter, up 363% year over year. Wall Street has treated that number as proof the AI infrastructure buildout is real, and that Oracle is one of its biggest winners.

I flagged this same "big number, unexplained catch" pattern three days ago with Broadcom's $164 billion backlog. Oracle's version is bigger, and the catch underneath it is sharper.

Because roughly half of Oracle's promise traces back to one company. OpenAI, via a reported $300 billion compute deal.

So the real question isn't whether Oracle can sign enormous contracts. It clearly can. The question is what happens to a $638 billion number if the company backing half of it can't fully use, or fully pay for, what it signed.

Cloud infrastructure revenue grew 93% year over year last quarter, hitting $5.8 billion. Total cloud revenue reached $9.9 billion, up 47%.

That is not a one-quarter fluke. OCI growth went from 55%, to 68%, to 84%, to 93%, across four straight quarters this fiscal year.

Management guided fiscal 2027 total revenue toward $90 billion, roughly 27% to 29% growth. For a company Oracle's size, that is a reacceleration almost nobody expected two years ago.

Of the $638 billion backlog, 12% converts to revenue within 12 months, about $77 billion. Another 34% converts within three years, another $290 billion.

Wall Street still overwhelmingly believes this works. 28 of 32 analysts rate Oracle a buy, with an average price target near $265, implying more than 75% upside from current levels near $149.

Here is where the cross-examination turns. The Wall Street Journal reported that roughly $300 billion of Oracle's backlog traces to a single deal with OpenAI.

Not a diversified customer base. One company.

In April, reports surfaced that OpenAI had missed its own internal user and sales targets for 2026. That is precisely the customer Oracle is depending on to convert half its backlog into real cash.

Because a backlog only pays out if the customer behind it can actually afford what it signed up for. If OpenAI's growth stumbles, Oracle's $638 billion promise does not shrink gracefully. It shrinks all at once, from its single biggest piece.

S&P Global Ratings downgraded Oracle's credit to BBB-, one notch above junk, on July 9. The stated reasons were customer concentration and a projected $42 billion free cash flow deficit for fiscal 2027.

Oracle burned $23.7 billion in free cash flow last fiscal year, versus a deficit of just $394 million the year before. Capital expenditures hit $55.7 billion, up 162%.

To fund fiscal 2027, Oracle plans to raise $40 billion to $50 billion in fresh debt and equity, including a roughly 5% dilutive equity program.

Michael Burry has been short Oracle through January 2027 puts. He covered half the position in July, saying plainly he was "covering half because winning." He kept the other half on.

Oracle closed August at $149.12, and the chart tells a rougher story than the backlog headline suggests. This is a stock that traded above $345 at its peak and has spent the year clawing back from a 52-week low of $114.50, a decline of nearly 60% from the top before this year's partial recovery.

That kind of round trip does not happen to a stock the market fully trusts. It happens when investors keep re-pricing the same story every time new information lands, euphoric about the backlog one month, panicked about the cash burn the next, cautiously buying the dip again after that.

Right now Oracle sits roughly 57% below its all-time high while carrying an average analyst target near $265. That gap is either the buying opportunity of the AI cycle, or a market slowly working out that the target price hasn't caught up with the real risk yet.

The next real test lands around September 9, when Oracle is expected to report fiscal first-quarter results. Until then, this stock is trading on narrative, and narratives move fast in both directions.

🎯 Whether Oracle diversifies its customer base beyond OpenAI
Watch for new large AI compute deals with other customers, Meta, Anthropic, or a major enterprise buyer. Even a $20-30 billion deal with a second name would meaningfully reduce the single-customer risk. Silence on new logos over the next month would confirm the concentration problem is getting worse, not better.

💵 How the next capital raise gets structured
Oracle needs $40-50 billion in fresh capital this fiscal year. Watch whether that leans more toward debt or equity. A debt-heavy raise adds pressure on an already-downgraded credit profile, while a larger equity component dilutes shareholders but protects the balance sheet.

📞 What Oracle actually says about OpenAI on the September 9 call
Management has stayed vague about customer-level backlog exposure so far. Watch whether analysts push for specifics on the call, and whether Oracle answers directly or deflects. A direct answer, even an uncomfortable one, would be more bullish than continued silence.

🐻 Whether Burry adds back to his short or exits entirely
He covered half his position while it was working. Watch his next 13F filing for whether he closes the rest or reloads. Either move is a real signal about how a sophisticated bear is reading Oracle's next few months, not just noise.

📊 OCI growth deceleration, not just headline revenue
The market has rewarded Oracle for OCI growth accelerating every quarter. The first quarter that growth rate slows from 93%, even slightly, is the first real crack in the bull case. Watch that number specifically, not just the total revenue headline.

I've heard both sides, and here is my verdict.

Oracle's cloud business is genuinely accelerating. 93% growth in infrastructure revenue is not a number companies fake, and $77 billion of the backlog converts to real revenue within a year regardless of what happens with any single customer.

But I can't get past the concentration. Not because OpenAI is a bad company. Because betting half of a $638 billion backlog on one customer's ability to keep growing is not diversified risk, it is concentrated risk wearing a diversified number's clothing.

Michael Burry did not cover his entire short. He covered the half that already worked and kept the half betting the story isn't finished. That is worth sitting with.

My actual read: Oracle survives this. The bigger question is whether it survives as the same company, or as one that had to dilute shareholders and restructure customer risk to get there. I would not add to this position until the September 9 call gives a straight answer on OpenAI, because right now the bull case is betting on someone else's balance sheet holding up.

Creators lock in holiday content calendars 90 days out, before brands figure out commissions. Waiting too long to launch an affiliate program means less runway to build demand and a missed shot at the best partnerships.

The 90-Day Holiday Sprint covers commissions, recruiting, and scaling a program at Day 30, 60, and 90.

Download the Free Guide

The best investing lesson from Oracle right now has nothing to do with AI. It's about concentration.

A big number feels safe. Six hundred thirty eight billion dollars sounds like diversification by size alone, when it is actually the opposite if half of it comes from one signature.

The same math applies to any portfolio, any business, any backlog. Size is not the same thing as safety.

The companies that get hurt worst in a downturn are rarely the ones with small numbers. They are the ones whose big numbers turn out to depend on one customer, one market, or one assumption holding steady.

Ask what a number is actually made of before you trust its size.

If this changed how you look at Oracle's backlog, send it to someone still repeating the $638 billion headline without asking who's behind it.