To The Moon

$Oracle(ORCL)$  

Mathematical Money | October 11, 2026

$ORCL$ fell 5.5% on Wednesday and then took 4.2% of it back on Thursday, finishing the week down about half a percent. If you only looked at Friday's close you'd think nothing happened.

Something did happen, and it wasn't really about the number everyone reported. The reaction was pointing at a genuine problem and aiming at the wrong figure entirely.

What the news actually said

The Financial Times reported that OpenAI told investors its annualised revenue was approaching $50 billion at the end of September, against a $70 billion figure that had been circulating. A $20 billion gap, roughly 30% of the headline.

Here's the part that got lost in the reaction. That gap wasn't lost business. It came from OpenAI changing how it annualises, so the figure could be compared like for like against Anthropic's. Same company, same customers, same contracts, different arithmetic. Nobody cancelled anything.

So the market sold a definitional change, then spent Thursday buying it back. That's the bit that looks like an overreaction, and on the surface it was.

I think the reaction was pointing at something real anyway. It was just aimed badly.

The number that actually matters

Oracle's remaining performance obligations — contracted revenue it expects to collect but hasn't yet, stand at roughly $664 billion. Against annual revenue guided to around $90 billion, that's a backlog more than seven times the size of the business.

OpenAI is about half of it. More than $300 billion of that contracted revenue comes from one customer.

Two things from Oracle's own earnings deck are worth putting next to that. The backlog grew $209 billion year on year but only $26 billion quarter on quarter, so the rate of growth has slowed sharply. And on the same slide Oracle lists "acceleration in customer diversification" as a highlight, which is the company telling investors directly that it knows the concentration is the issue and is working on it. I'd take that seriously rather than dismiss it, and I'd also note that a single quarter's diversification does not move a number this size.

Now put the two numbers side by side, which is the thing almost nobody did this week. OpenAI's entire annual revenue run rate is about $50 billion. Its contractual commitment to Oracle alone is more than six times that, and that's before it pays for anything else it needs, including the chips, the people and the research.

That's the actual risk, and it did not change on Wednesday. It was exactly as true the week before. What the revision did was make people look directly at a number they'd been carrying without examining.

Oracle's half of the problem

The customer is funding a commitment larger than its revenue. The supplier is funding the build-out the same way.

Oracle's fiscal 2026 capital expenditure came in at $55.7 billion. Operating cash flow was a record $32.0 billion and was comprehensively swamped by it, leaving free cash flow at negative $23.7 billion. The company already carries more than $100 billion of debt and has signalled another $45 to $50 billion of raising this calendar year to keep building.

S&P cut Oracle's credit rating from BBB to BBB− back in July, and the reason they gave was precisely this customer concentration. That downgrade is three months old and I'd suggest it was better information than anything that moved the stock this week.

So you have a supplier borrowing heavily to build capacity for a customer that has promised it more than six times the customer's current annual revenue. Both sides of that are funded by expectation rather than by cash flow. That can absolutely work. It is also a materially different proposition from a backlog, which is how it usually gets described.

The chart, honestly

Oracle closed Friday at $141.40. It trades below its 20-day, 50-day and 200-day averages, with the 200-day about 12.6% above the price.

It is 54.8% below its 52-week high of $313, set almost exactly a year ago. Sixty-day volatility runs near 54%, so a 10% week is ordinary here rather than dramatic.

This is not a stock that fell on Wednesday's news. It's a stock that halved over a year and had a bad Wednesday.

Where I am

I hold October 2027 calls at the $120 strike, deep in the money, bought after I was called away from the shares earlier in the autumn. They're down about 13%. Against them I've written short calls at $180 expiring in early November.

Here's the honest arithmetic on it, because this is the part that decides everything else. With the stock at $141.40 and a $120 strike, a little over half of what those calls are currently worth is time premium rather than intrinsic value. That premium goes to zero by October 2027 whether the stock moves or not. For the position to break even I need Oracle somewhere around $170 by then — roughly 20% above where it trades now, and conveniently just above where the 200-day average currently sits.

That's the real question in front of me. Not whether I like Oracle, but whether it gets 20% higher inside twelve months.

The plan, in three rules

One: I don't write calls below $170. That's the break-even on the long position, and writing below it means capping the stock exactly where I need it to go. It's the same rule I apply to not writing calls under my cost basis elsewhere, and it binds hardest when the premium below that level looks most tempting. The $180s I've written sit comfortably above it.

Two: I accept that the call writing only does part of the job. At the $180 strike the premium covers roughly 40% of the time decay I'm carrying each month. I could cover all of it by writing much closer to the money, and that would break rule one. So the position needs the stock to do some work. Pretending otherwise by selling nearer strikes would feel productive and would quietly convert a directional bet into a capped one.

Three: the trigger is the backlog and the funding, not the price. I'm not going to react to another week like this one. What I'll actually watch is whether the RPO keeps growing, whether Oracle's borrowing stays on schedule, and whether OpenAI's funding keeps pace with what it has committed to spend. If the backlog stops growing while the capex keeps going, that's the signal, and it'll show up in a filing rather than on a chart.

And a fourth that's really a consequence of the first three: I'm not adding. The concentration is the thesis risk. Buying more of it because it got cheaper is just taking more of the same bet.

What would change my mind

Not the share price, and not another revenue revision at OpenAI. It would be Oracle's cash flow deficit widening while the backlog stops growing. Right now the enormous spending is justified by an enormous order book, and that argument holds as long as both are true. If the book stalls, the spending stops being investment and starts being exposure, and the credit rating becomes the story rather than a footnote.

The other one is simpler. If OpenAI raises on materially worse terms than the last round, the market is telling you something about the counterparty that Oracle's backlog can't tell you, because contracted revenue is only worth what the counterparty can pay.

Two things I'd like other views on.

How do you treat customer concentration when the customer is private? With a listed counterparty you can read the accounts. Here the thing that determines whether half of Oracle's backlog converts is a company that discloses what it chooses to, when it chooses to, and this week that included redefining its own revenue measure.

And on the structure: when you hold long-dated calls that need the stock meaningfully higher to break even, do you write against them at all? There's a decent argument that you shouldn't, because the premium is small relative to the move you need and every strike you sell is a ceiling. I've written them anyway. I'm not certain that's right.

Backlog, capex and cash flow figures are from Oracle's Q1 FY2027 earnings presentation, 10 September 2026. OpenAI's revenue run rate is as reported by the Financial Times, 8 October 2026. Prices are from Tiger daily bars.

#ORCL #OpenAI #AI #options #LEAPS

Stop guessing. Start calculating.

Live to fight another day. 🤙