This week started with a question about backlogs.

Wednesday it was Oracle, half its bookings tied to one customer, and the market punishing a stock for math it couldn't verify.

Even Nvidia spent $3.5 billion this week buying its way into a second chip partner, which tells you the biggest player in AI still isn't sure who wins this outright.

Thursday night, Broadcom got asked the backlog question directly, except this time with actual results instead of a promise.

AI revenue grew 221% year over year to $16.7 billion.

Guidance called for another 236% next quarter.

The stock dropped anyway.

Most people are filing that under "the market is never happy," the same lazy explanation that gets reached for every time a good number doesn't move a stock.

I think that read misses what actually happened this week, and it says something uncomfortable about how high the bar has gotten for anything tagged AI.

That is what The Pragmatic Playbook is built around today.

Missed Wednesday's deep dive on Oracle's backlog problem?

Google is trying to flip the narrative after its longest monthly losing streak in more than a decade. The stock has been pressured by AI capex worries, antitrust overhangs, and questions over whether search can stay dominant in an AI-first world, but recent court news looks less damaging than feared and investors are giving Alphabet more credit for its AI and cloud positioning. In plain English, Google is still carrying baggage, but the AI story is starting to regain momentum.

Classic Buffett: market crashes feel awful, but they are often when long-term investors get their best chances. The article leans on Buffett’s famous idea that “bad news is an investor’s best friend,” because lower prices let patient investors buy good businesses at better valuations. The takeaway is not to predict the crash perfectly, but to keep cash, discipline, and a watchlist ready when fear returns.

Qualcomm is making real progress beyond handsets, especially in automotive, IoT, and data-center AI, but the smartphone business is still too big to ignore. Automotive and AI growth could help offset weaker handset demand, but investors want proof that these newer segments can scale fast enough to change the story. In plain English, Qualcomm has a credible diversification plan, but the market is not ready to forget its smartphone exposure yet.

"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

This has been a strange week for anyone who owns AI infrastructure stocks.

Oracle reported a backlog north of half a trillion dollars on Wednesday and got sold anyway, because half of it traces back to a single customer nobody outside the company can independently verify.

Nvidia put $3.5 billion into MediaTek this week to keep expanding its own AI ecosystem, a reminder that even the biggest player in the room is still hedging its bets.

Then Broadcom reported Thursday night, and for a moment it looked like the exception to a rough week for AI names.

Revenue beat. Earnings beat. AI semiconductor revenue grew 221% year over year to $16.7 billion.

The easy read is that Wall Street has simply lost the plot, that no number will ever be enough, and that this is what a bubble looks like from the inside.

I don't think that's it.

Pull the guidance apart and a narrower, more useful story shows up.

Broadcom didn't miss. It beat by less than the market needed it to.

That is a different problem, and it is the one worth understanding before you decide what to do with this stock.

Broadcom's AI semiconductor revenue hit $16.7 billion in the quarter, up 221% year over year and 54% from the prior quarter alone.

That segment carried a Semiconductor Solutions unit that did $20.8 billion in total, up 127% year over year.

Total revenue came in at $29.6 billion, ahead of the $29.4 billion Wall Street wanted, and adjusted earnings of $3.32 a share beat the $3.23 estimate.

Free cash flow was $13.7 billion for the quarter, up 95% year over year and representing 46% of revenue.

None of that reads like a company running out of demand.

Because this is the exact question Oracle's backlog raised two days earlier: whether AI spending is actually converting into cash, or just sitting on a spreadsheet as a promise.

Broadcom's answer this week was concrete numbers instead of a projection.

Here's the part that should bother you more than it seems to be bothering the stock price.

Guidance for next quarter came in at $34.8 billion.

Wall Street wanted $35.05 billion.

That gap is less than 1%, and it still knocked the stock down as much as 6% in after-hours trading.

Not because the number was bad.

Because the number wasn't extraordinary enough for what this stock now demands of it.

StoneX analyst Cody Acree, who still rates the stock a buy, put it plainly: "the magnitude is not quite enough from a top and bottom line standpoint on the beat and raise when you have a company that is this levered to AI."

That is the real story of this earnings report.

Broadcom is up only 6% year to date, while Nvidia is up 20%, and the market has decided that being merely excellent isn't the same as being unstoppable.

Snowflake also reported this week and jumped 22% on its beat.

That's the tell. This market hasn't stopped rewarding good numbers. It just decided Broadcom's needed to be extraordinary.

This is where Wednesday's Oracle story and Thursday's Broadcom story rhyme.

Back in August, Broadcom disclosed $164.6 billion in firmly committed AI orders and still wouldn't say which customers make up that number, or how concentrated it is. I covered that gap in detail here.

That silence didn't get any louder this week, even with a full earnings call available to fill it.

Layer on Marvell, which landed a custom AI chip deal with Google in August that came with an option for Google to buy a $12.2 billion stake in Marvell.

That single deal ended Broadcom's run as the default hyperscaler chip partner.

Goldman Sachs has argued the market is overreacting to the competitive threat.

Maybe. But maybe is not the same as an answer, and Broadcom still hasn't given one.

Broadcom spent the week sitting almost exactly on top of both its 50-day and 200-day moving averages, a stock that couldn't decide which way to lean into earnings.

Thursday night decided for it.

The stock gapped down as much as 6% in after-hours trading and settled the session down closer to 4%, back near $354.

That drops it below both of those moving averages for the first time in weeks, and it puts the stock roughly 28% off its 52-week high of $495, right back in the same zone it was sitting in when I wrote about the $164 billion backlog at the end of August.

The level worth watching into next week sits around $340 to $345.

That's where buyers showed up the last time this stock got this uncertain about itself, and if it holds again, it tells you the backlog thesis is still intact even if the market's patience for the story is wearing thin.

Management still hasn't broken out customer concentration inside that AI backlog, and this earnings call was another chance to do it that came and went.

Bulls will tell you this is normal, that hyperscaler contracts always come with confidentiality clauses attached.

Bears will point to Oracle and say the market is done taking that explanation on faith.

The wildcard: Broadcom doesn't have to disclose this voluntarily. A regulatory filing or a customer's own earnings call could do it for them, on a timeline it won't control.

Marvell's Google deal is real, and it broke Broadcom's run as the only serious custom silicon option for the largest AI spenders.

Bulls say the market is big enough for two winners, and Broadcom's scale still makes it the safer bet.

Bears say every hyperscaler now has a credible second source, and pricing power starts eroding from here.

The wildcard is whether another hyperscaler follows Google's lead before Broadcom reports again.

The backlog only means something if the customers behind it can pay for it without leaning on debt that looks shaky the moment growth slows.

Oracle's stock got hit on exactly this question this week.

Any crack in how Broadcom's hyperscaler customers finance their AI capex will hit this stock before it hits the fundamentals.

That range held the last time sentiment on this stock got shaky, back in late August.

A clean hold into next week says institutional money still believes the backlog story despite a noisy guidance miss.

A break below it says the market is starting to price in real doubt, not just short-term disappointment.

I went into this earnings report expecting Broadcom to either prove the AI story or expose it.

It did neither, cleanly.

What it actually did was more interesting. It proved the demand is real, and it showed you, in the same breath, that "real" isn't the bar anymore.

221% revenue growth used to be an obviously good number. This week, it wasn't good enough to hold the stock up.

That's not a Broadcom problem. That's a market that has quietly reset its definition of a disappointing quarter for anything wearing an AI label.

The lesson I'm carrying into the weekend is that the backlog was never really the risk here. The multiple was.

Broadcom trades at more than 60 times earnings, and at that price, excellent needs to come with a surprise attached, not just a beat.

Watch what the market rewards over the next few weeks, not what it praises in headlines.

Right now, the market is telling you exactly how expensive perfection has become.

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

This week, two AI infrastructure stocks got hit for two different reasons, and neither reason was "the demand disappeared."

Oracle got questioned on the quality of its backlog. Broadcom got questioned on the size of its beat.

Neither company failed. The market's expectations just kept moving faster than the results could catch them.

That's worth sitting with before Monday.

The setups that survive weeks like this aren't the ones chasing the loudest headline number. They're the ones built on a level you'd still trust once the excitement wears off.

Know your support levels before earnings season gets loud again next week, not after.

The best readers here were referred by readers like you. Pass it on.