This week had a theme, even if nobody put a name on it.

Broadcom turned itself into a lender for its own customers. Intel watched its biggest proof point start quietly shopping other chip partners, a strange thing to watch happen to a relationship everyone assumed was locked down. Then on Tuesday, Marvell topped them both.

Three different tickers, one uncomfortable question. When a chip company has to pay customers to stay loyal, is that strength, or is that fear wearing a nice suit?

Today's Pragmatic Playbook digs into Marvell's wild week. A new $20 billion target, a first-ever $90 billion dream for 2031, and a warrant buried in the fine print that tells you more than either number does.

Missed Wednesday's deep dive on Intel?

The key issue is whether OpenAI can grow fast enough to justify the enormous AI infrastructure spending being built around it. Reports of weaker-than-expected revenue figures pressured Nvidia and other AI-linked names because investors immediately questioned whether current data-center and compute commitments are running ahead of monetization. In plain English, the AI boom is still huge, but the market is starting to demand proof that revenue can keep pace with capex.

Starbucks-Chipotle deal would effectively reunite Brian Niccol with the company he previously transformed. Chipotle shares jumped on reports Starbucks has discussed a possible bid, but no formal offer has been confirmed and a deal of this size would be extremely complex. The opportunity is obvious, with shared real estate, loyalty data, and operating know-how, but the risk is that Starbucks could distract itself just as its own turnaround is gaining traction.

Alphabet has spent nearly five months building a tight technical base while the underlying business keeps improving. Google Cloud growth is accelerating, Search remains resilient, profitability is strong, and the stock still trades at a relatively modest valuation compared with other mega-cap AI names. With analysts seeing roughly 22% upside, the setup is interesting because the technical and fundamental stories are finally pointing in the same direction.

Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:

After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.

The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.

“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”

Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.

Learn More

This week, three chip companies tried to buy certainty instead of earning it.

Broadcom financed its own customer's purchases. Intel lost its only named chip partner to a rival's fab talks. Then Marvell raised the stakes on both of them at its investor day.

The easy read is that Marvell just had the best week of its year. A new $20 billion target for fiscal 2028, and for the first time ever, a number on 2031: $70 to $90 billion. Shares jumped 10% Tuesday, then another 5.8% Wednesday.

Pull that apart and it gets more interesting.

Buried in the investor day materials is a warrant giving Google the right to buy up to 59 million Marvell shares, worth roughly $12.2 billion at the strike price. Google only earns that stock by buying more Marvell chips.

That is not a customer win. That is a dowry.

The Pragmatic Playbook today isn't about whether Marvell's numbers are big. They are. It's about whether a growth story partly funded by equity handed to your own customer deserves the same multiple as one built on orders you simply won.

Marvell didn't just beat expectations this week. It rewrote them.

The fiscal 2028 revenue target moved to $20 billion, up from $18 billion. That's above the roughly $18.2 billion Wall Street was modeling going in.

For fiscal 2029, the custom silicon target rose to more than $12 billion, up from more than $10 billion.

And for the first time, management put a number on fiscal 2031 at all. $70 to $90 billion, against a Wall Street consensus heading into the week of just $47 billion.

That's not a beat. That's Marvell telling analysts their whole model was off by roughly half.

CEO Matt Murphy called a $30 billion custom chip target "not a stretch." He floated the idea of Marvell becoming a trillion-dollar company out loud, on stage, in front of investors still digesting last quarter's numbers.

Here's where it gets uncomfortable.

That "$120 billion" figure everyone is repeating isn't a sale. It's a ceiling.

Google's warrant vests in 240 tranches of $500 million each, tied strictly to actual chip purchases. The window runs from fiscal 2027 through fiscal 2033, and there's no minimum anywhere in that structure.

Google buys the chips, or it doesn't earn the stock. Nothing obligates it either way, and Google is simultaneously buying custom silicon from Broadcom and from MediaTek.

Marvell didn't win a customer this week. It rented one, and paid for the lease in its own equity.

To hit even the low end of that $70 billion 2031 target, Marvell needs roughly 49% to 57% compounded annual growth. Sustained for five straight years.

No large-cap semiconductor company has done that in the modern era. Marvell's current pace sits closer to 36% to 37%.

The valuation already assumes the dream comes true. Shares trade near 96 times trailing earnings, more than double the five-year median.

And the people who know the company best haven't been buying. Insiders sold roughly $52 million in Marvell stock over the past 12 months. They bought zero.

Goldman Sachs isn't buying the $90 billion story either. It holds a Neutral rating with a $220 target, well below where the stock trades heading into the weekend.

Rewind five weeks and Marvell looked like a different stock entirely.

Before its August earnings, shares traded near $330, a 52-week high. The quarter's guidance disappointed, and the stock fell roughly 10% in a day, bottoming in the low $220s by early September.

I flagged that bottom as the setup worth watching back in Marvell Beat The Quarter. Wall Street Sold It Anyway., when the average analyst target sat right around $290.

Five weeks later, the stock has clawed almost that entire round trip back. Tuesday's 10% jump pushed shares straight through their 50-day moving average. Wednesday added another 5.8%, touching $287.

That puts Marvell almost exactly back at the target analysts were using a month ago. Except the targets themselves have moved too, with the average 12-month number now sitting near $305.

Thursday brought a small pullback, about 1.6%, which traders are calling ordinary profit-taking after two enormous up days.

The level that matters heading into next week is Goldman's $220. That's roughly where this stock bottomed in September, and it's the number the Street's most prominent skeptic is using as fair value today.

The warrant means nothing until purchases happen. Watch Marvell's fiscal Q3 2027 commentary, the first window where tranches can start vesting. Bull case: early vesting signals real demand, not paper promises. Bear case: Google's own TPU roadmap could shift and the ramp slides right along with it. Wildcard: Google leans harder on Broadcom or MediaTek, and the warrant simply expires mostly unused.

Piper Sandler and BofA raised targets this week. Goldman didn't move an inch. Bull case: more firms converge toward $305 or higher, confirming the story. Bear case: a split this wide, with Goldman nearly a quarter below consensus, usually means the stock is pricing in an argument, not a conclusion. Wildcard: six of 46 analysts already rate it Hold, more skepticism than you'd expect after a week like this one.

Every number through 2031 depends on hitting $20 billion in fiscal 2028 first. Bull case: Murphy has a real track record of raising guidance and then beating it. Bear case: one soft quarter here doesn't just miss a number, it undercuts every target stacked on top of it. Wildcard: supply constraints at TSMC's advanced nodes could cap everyone's growth at once, Marvell included.

Amazon is reportedly dual-sourcing its Trainium chips rather than leaning on one vendor. Bull case: hyperscalers still need Marvell's design expertise even while hedging. Bear case: if every major customer starts splitting orders the way Amazon reportedly is, Marvell's multiple-hyperscaler story gets thinner at the edges. Wildcard: a visible Amazon pullback would be the clearest test yet of how real this demand actually is.

Right now the signal is one-sided. Bull case: executives buying here would be the strongest vote of confidence available. Bear case: continued selling with zero buying doesn't inspire trust in a stock already priced for perfection. Wildcard: an insider purchase announced alongside Q3 2027 results would line up perfectly with the warrant's first vesting window.

I've spent this whole week watching chip companies try to buy the thing they used to just earn.

Broadcom financed its own customer's purchases. Intel lost its only named chip partner to a rival's fab. Marvell handed Google twelve billion dollars in stock just to keep the orders flowing.

Not because the chips got better. Because the contracts got more creative.

And that makes Marvell a very different company to own than the one most people think they're buying this week.

Because when the biggest number in your investor deck depends on a legal contract instead of a better product, the contract is the real story. Not the headline revenue target sitting above it.

Marvell's $90 billion dream might come true. Matt Murphy has earned some benefit of the doubt here.

But I'm not grading this stock on the number on the slide. I'm grading it on how much of that number had to be bought instead of won.

This week taught me the biggest headline growth target isn't the one worth watching. The size of the check a company had to write to get there is.

Wispr Flow turns your voice into clean, professional text inside any app. Emails, Slack, client updates: speak once, send without editing. 4x faster than typing. Works on Mac, Windows, iPhone, and Android.

Try Wispr Flow free

This week rewarded anyone who read past the headline number.

Marvell's chart moved fast. The warrant attached to it moved just as fast, quietly, in the background.

That's usually how the real risk travels.

Big promises feel good heading into a weekend. Fine print is what decides whether you're still comfortable holding on Monday.

Before you chase anything this week told you to buy, ask what the company had to give up to tell you that story.

Then ask whether you'd still want the stock if that number never shows up at all.

Know someone who'd rather think than chase? Forward them this edition.