Dear Income traders,

Two companies posted monster quarters this week. One got sold 19 percent in a single session. The other jumped 10 percent.

Neither reaction had much to do with the quarter.

When a tape gets this crowded, the print stops being the story and positioning becomes the story.

One of these handed you a gift. The other handed you a bill.

Somebody Has to Instrument the Pipes

Datadog (DDOG) did $1.12 billion, up 36 percent year over year against a $1.08 billion estimate, with EPS of 65 cents against 58 cents. Management raised the full year to $4.45 to $4.47 billion, up from the $4.30 to $4.34 billion they guided to in May. Next quarter's midpoint also came in above the Street.

Then the stock fell 19 percent.

Here's what actually did it, and it wasn't the raise. Datadog's largest customer, a nine-figure AI account running 17 of its products, told them it's cutting usage starting in the third quarter. Guggenheim moved the name to Sell with a $105 target on the view that this customer is building observability in-house.

That's a real risk. I'm not going to pretend it away.

But take that one account out and look at what's underneath. Revenue growth from the non-AI customer base accelerated to the high 20s, up from the mid-20s last quarter and 18 percent in the year-ago quarter. That's the boring part of the business speeding up, not slowing down.

The platform numbers say the same thing. Customers running four or more Datadog products went to 58 percent from 52 percent a year ago, and the ones running 10 or more nearly doubled to 13 percent. Clients paying $100,000 a year or more rose 23 percent to about 4,720.

That's not tire kickers. That's AI workloads showing up in the receipts.

Why I want it: data observability (the software that watches whether your systems are actually working) is the toll booth on compute, and the toll gets paid more as compute gets cheaper, not less. Yeah, I know. Hans, cheaper models and Chinese open source mean less spend.

No. Jevons.

When inference gets cheap you don't run fewer workloads. You run 10 times as many, and every one of them has to be watched, secured and billed. The pipes multiply, and somebody has to instrument the pipes.

At roughly $237, Datadog trades near 20 times the midpoint of this year's revenue guide. Still not cheap, and a name at that multiple can overshoot. But this is the kind of selloff you scale into over weeks, not the kind you chase in an afternoon.

I'm not calling the bottom. It's on my radar.

A Record Quarter That Couldn't Reclaim June

Nothing was wrong with Caterpillar's (CAT) quarter. That's the point.

Caterpillar printed its first $20 billion quarter on Tuesday: $20.54 billion, up 24 percent, with adjusted EPS of $8.17 against a $6.20 estimate and operating margin of 20.9 percent from 17.3 percent. Backlog jumped 92 percent year over year to $72 billion. Management raised the full year on top of it.

The stock popped about 10 percent, its biggest session since last October.

And it's still trading near $857, roughly 20 percent below the $1,062.93 it closed at on June 30.

Sit with that for a second. Best quarter in company history, record backlog, raised guide, and the shares can't get back to where they traded five weeks ago. July took 23 percent out of this stock, its worst month since 2009, and one blowout print bought back less than half of it.

That tells you something about who's left to buy up here.

The wrinkles are real, too. Management flagged dealer inventory in Construction Industries as a volume headwind in the back half, with the usual fourth-quarter drawdown stacked on top. Tariffs still run about $2.2 billion for the full year.

Then there's the long end. The 30-year sits near 5.2 percent, the highest it's been since 2007. Yeah, I know. Hans, nobody buys a bulldozer off the 30-year.

Fair. But this is a financing business: dealer floor plans, fleet purchases, mining capex, long-cycle projects, all penciled against the long end. Do deals get cancelled when the math tightens? Not really. Does the marginal buyer of a fleet or a mine think twice? A little.

And "a little" is plenty at 28 times forward earnings against a decade average closer to 18.

Great company. Wrong spot on the chart. Let somebody else have this one up here.

Which One Makes Monday's Cut

That's two of the 10 names hitting the table. The rest of the crew is bringing theirs, so what will the crew choose for the Ticker Highlight pick of the week?

Tap this link and pick your subscription rate and get Monday's pick as soon as it hits.

Here for a good time AND a long time,

Hans