To The Moon

Nvidia Scores +10%. Microsoft Scores +0.3%. Here Are Five That Beat Them Both.

Mathematical Money | September 25, 2026

Thursday looked like nothing happened.

S&P closed 7,704.13, down 1.9 points. Nasdaq 26,939.37, up 3.34. The Dow did the most work and it still only lost 161 points to 51,349.98. Call the whole session flat and you wouldn't be wrong.

But the 10-year touched 5.15% before settling back near 5.1%, and that's the highest it's been since 2007.

So the index did nothing while the rate everything gets discounted at went to a nineteen-year high. Those two don't sit together comfortably, and when my screens disagree with the tape I've learned to believe the screens.

Rather than write another recap of a day that didn't move, let me show you what I actually run.

The five things on my dashboard

Five macro inputs, each just compared to its 200-day average. Nothing clever. I'd rather have five signals I can't argue with than one opinion I can talk myself into.

Bitcoin $81,316 vs a 200-day of $70,463 — bullish

SPY $761.69 vs $712.48 — bullish

VIX 14.81 vs 18.09, and under 20 — bullish

Dollar index 100.22 vs 99.15 — bullish

10-year at 5.00% vs a 200-day of 4.40%, above my 4.25% line — bearish

Four green, one red. Risk-on, one warning light.

And the warning light is the thing that moved on Thursday. That's why a flat day is worth writing about.

I don't use price targets

Never have, really. A single target pretends to a precision nobody has, and I've watched too many analysts quietly move theirs after the fact.

What I run instead is a probability-weighted expected value. Three scenarios, a probability on each, multiply and add. The maths is secondary school level — it's being honest about the probabilities that's hard.

Bull case, with a range and a likelihood. Base case, same. Bear case, same.

What comes out isn't "this is going to $600." It's "across everything that might plausibly happen, weighted by how likely I reckon each one is, here's what I'm being paid to take the risk."

The useful part is that the number can come out strongly positive on a company with genuinely ugly headlines, and it can come out flat on something everyone loves.

One thing before the list. I track how far each stock has moved since I last wrote its thesis, because a good number on a stale price is just a wrong number wearing a suit. I pulled one name out of this post entirely this morning for that reason — it had run 12% since the screen and anything I said about it would have been out of date. So each name below comes with its screen date and drift, and you can discount accordingly.

1. AppLovin (APP) — +48.9%

Screened 19 Sep, price has drifted +1.1% since. Fresh.

Around $311, roughly $141 billion market cap, down about 41% this year.

Bull $550-$650, 35% likely

Base $380-$480, 45%

Bear $250-$350, 20%

The numbers don't look like a stock that's halved. Q2 revenue was $1.92 billion, up 53% year on year. Adjusted EBITDA $1.61 billion on an 84% margin. Q3 guidance implies 46-48% growth.

So why's it down 41%? It missed consensus by twenty million dollars, which in a crowded momentum name is plenty. And there's still a short report from March 2025 hanging over it alleging inflated ad returns — no regulatory action since, but the market hasn't forgotten either.

The thing I'm watching is AXON 2.0, the self-serve product they launched in June, which opens up e-commerce advertising. Early returns are running at parity with Meta's network. Q3 results in November will tell us whether that's real.

Look at the bear case though. Even down at $300 I'm only losing about 3%, because the multiple already collapsed from 68 times EBITDA to 33. Most of the bad news is in the price. That's the asymmetry, and it's why this sits top of my list.

2. Royal Caribbean (RCL) — +34.9%

Screened 12 Sep, price has drifted −13.2% since. Ranges are two weeks old at a higher price — treat as directional.

Around $225. About $77 billion.

Bull $320-$370, 35%

Base $275-$310, 50%

Bear $215-$270, 15%

Two earnings beats back to back. Q1 EPS $3.60, twelve percent ahead. Q2 $4.21, seven percent ahead. Full-year guidance of $17.73-$17.87 is fourteen percent growth, capacity is up 6.7%, pricing is holding.

Then it fell 9.7% over ten sessions, for no reason I can find.

Consensus target is $346.92, about a third above where it trades. The risks are at least honest ones — fuel costs them 62 cents of EPS and that's already baked into guidance, and if the consumer genuinely rolls over, cruises are one of the first things people stop booking.

The most boring business on my list, which is rather the appeal. No AI angle, no story, just a company beating estimates whose shares went down anyway.

3. Broadcom (AVGO) — +34.8%

Screened 24 Sep, drift +0.2%. As fresh as it gets.

Around $358.

Bull $520-$580, 45%

Base $440-$500, 40%

Bear $280-$350, 15%

Their September quarter was silly. Revenue $29.6 billion, up 86%. AI revenue alone was $16.7 billion, up 221% — that's now more than half the company. Next quarter they've guided to $34.8 billion, and management say $350 billion of shipments are already locked in for FY27 and FY28.

Here's what worries me. Top five customers are 55% of quarterly revenue now, up from 40% a year ago. Six hyperscalers basically decide what this company earns. One of them slows capex and the bear case turns up very quickly.

Highest bull probability on the list at 45%, and the best argument I know for sizing something carefully rather than enthusiastically.

4. Chipotle (CMG) — +28.9%

Screened 12 Sep, drift −10.1%. Same caveat as Royal Caribbean.

Around $32.50 after the split. Roughly $45 billion.

Bull $45-$52, 35%

Base $38-$43, 50%

Bear $28-$34, 15%

A margin story rather than a growth story. Comps were only 2.2% last quarter and revenue grew 9.3%, but operating margin fell to 15.7% from 18.2%. Beef cost them 80 basis points, labour another 30.

Underneath that squeeze they're still building. 350-370 new restaurants this year, 80% of them the drive-through Chipotlane format which does higher volumes. Digital is 38.3% of sales now. And they opened Seoul on the 2nd of September, their first real step into Asia — that's the bit I find most interesting and also the bit least likely to matter within twelve months.

Q3 earnings land 28 October with about 1% comps guided. Beef and freight should ease by Q4. If they do, the worst of the squeeze is behind them.

5. Take-Two (TTWO) — +27.0%

Screened 12 Sep, drift −5.7%.

Around $205. About $36.5 billion.

Bull $270-$320, 45%

Base $225-$265, 35%

Bear $190-$225, 20%

GTA VI launches 19 November. Biggest entertainment launch anyone's attempted, and it drops straight into the holiday quarter.

Q2 revenue beat by 11.5% but full-year guidance came in 18.3% light, which is exactly what you'd do if you were managing expectations ahead of something this size. Stock's down about 10% this year. 28 of 29 analysts rate it a buy, which is either conviction or herding depending on your mood.

The bear case isn't the game flopping. It's that everyone already knows it'll sell — record launch day, then everyone takes profit. That's why the bear range only gets down to $190, a 4% drawdown, while the bull runs 36% higher.

Now the uncomfortable bit

Same model, same method, applied to the names on everybody's watchlist. I've only included ones screened recently where the price hasn't run away from the thesis.

Alphabet +15.9%

Amazon +11.2%

Mastercard +11.0%

Nvidia +10.5%

Visa +10.1%

Lilly +7.6%

Apple +6.7%

PayPal +5.4%

Microsoft +0.3%

Coca-Cola −0.3%, avoid

Microsoft comes out at a rounding error over twelve months. Nvidia, the most-discussed stock on earth, scores less than a quarter of what AppLovin does.

I want to be careful here, because that's not me saying these are bad companies. They're excellent companies, obviously. What it says is that at these prices, with ranges I can actually defend, you're not being paid much to hold them for the next year. The market already agreed with you about the quality — that agreement is what the price is.

The five above aren't better businesses than Microsoft. Not close. They're businesses where something's recently gone wrong, or where everyone's waiting for proof before they'll pay up. That's generally where the expected value hides, and it's also exactly why it feels uncomfortable to own them.

What I actually hold

Of those five I own long-dated calls on AppLovin and Broadcom. Not Royal Caribbean, not Chipotle, not Take-Two. They rank well and I haven't acted — which I mention because a list of names someone is already long is worth less than one where they tell you what they've skipped.

On the rate move, I'm doing nothing. My long-dated positions run to 2027 and 2028, and I bought that much time specifically so a yield scare wouldn't force my hand. If 5.15% on the ten-year breaks a thesis, the thesis was never really about the company.

What would change my mind is the yield staying up here into year-end instead of spiking and fading. A 5.1% that sticks re-rates growth multiples for good. One that pokes up and comes back down is just a scare. Very different things.

Two things I'd genuinely like to hear from you on.

First — does anyone else write down actual probabilities before they buy? Putting a number on "20% chance I lose 3%" is honestly the single habit that stopped me sizing like an idiot. I don't know whether most people do this or whether I'm just being obsessive.

Second, and I expect to get argued with: my model says Microsoft returns roughly nothing over the next twelve months on a probability-weighted basis. Tell me where that's wrong. Seriously — if you think my bear range is too harsh or my bull probability too low, say so and I'll re-run it and post whatever comes out, even if it makes me look silly.

Stop guessing. Start calculating.

Live to fight another day. 🤙

Nothing here is financial advice. These are my own scenario estimates from my own framework, and reasonable people will land on different probabilities. I hold long-dated calls on APP and AVGO. Do your own work.