For most of this year, the market’s running belief has been simple: AI infrastructure deals are automatic wins.

Scale up. Lock in the picks-and-shovels story. Let the stock take care of itself.

Marvell just put that belief on trial.

On Monday, it announced an all-stock, $115 billion deal for Lumentum, built to dominate AI optical connectivity before anyone else could get there first. Marvell’s stock fell almost 13%. Lumentum’s fell almost 18%.

That is not what an “automatic win” looks like the morning after.

This week’s focus is Marvell, and a number buried in the deal math that nobody’s mentioning on CNBC: by the time Monday’s selling stopped, the premium in this “premium” deal had already flipped into a discount.

Missed Monday’s edition on Meta’s widening credit gap? 🤔 Meta Beat Earnings. Why Is Its Credit Rating Falling?

Apple delivered fiscal Q4 revenue of $114.5 billion and EPS of $2.10, helped by strong demand for the iPhone 18 Pro and record Services revenue. The catch is margins: Apple expects December-quarter gross margin to fall to 45% to 46% as memory-chip costs rise, while Services growth is also expected to slow. The bigger takeaway is that Apple’s demand remains healthy, but component inflation could become the next major pressure point for earnings.

Palantir’s Q3 revenue jumped 96% to roughly $2.31 billion, while U.S. revenue surged 119% and commercial revenue surpassed government revenue for the first time. That is important because it shows Palantir is becoming much more than a government contractor, with AI adoption spreading rapidly across private companies. The caution is that new contract signings barely grew, so investors will want to see whether today’s explosive revenue growth can be sustained.

Starbucks posted 7.7% U.S. comparable-sales growth and adjusted EPS of $0.80, while margins improved sharply and management said its “Back to Starbucks” turnaround has become a growth plan. The catch is that transaction growth slowed to 2.4%, with higher spending per customer doing more of the work. In plain English, Starbucks has clearly stabilized, but the next phase depends on getting more people through the doors rather than simply charging existing customers more.

There's a new channel of revenue that has just been unlocked for companies with over 20 people that have been around longer than a year. 

Few know it, but AI labs now let you monetize your company’s operational data, and they're spending $15 billion this year on it to train the next frontier of LLMs. Your workflows, documentation, processes, and internal convos like Slack are extremely valuable and they're paying anywhere from $100k to $3 million for it per company.

Before you ask, all the data is anonymized and cleaned before it's sent to a Lab. Bezel (Y combinator backed) helps find a data partner that is the best fit for you.

Check your data's worth

Marvell says this deal makes it the backbone of AI’s next bottleneck: optical connectivity.

That’s the pitch. Marvell is paying $115 billion in its own stock for Lumentum, and both companies are calling it a category-defining move into the physical pipes that move data between AI chips.

On paper, the logic is clean.

Marvell builds the custom silicon inside hyperscaler AI racks. Lumentum builds the optical components that carry data between those racks at the speeds modern AI clusters demand.

Combine them, and Marvell stops competing for sockets one customer at a time. It starts owning both ends of the connection.

The market’s problem is that it didn’t buy this story for even one trading session.

Marvell dropped nearly 13% the day the deal was announced. Lumentum, the company supposedly getting acquired at a premium, dropped almost 18% right alongside it.

Normally, when a target agrees to be acquired, its stock drifts up toward the announced deal price as arbitrage traders lock in the spread.

Instead, Lumentum’s stock cratered in the same direction as its acquirer’s, on the same day, for the same reason.

That is not how arbitrage is supposed to work.

So the real question isn’t whether AI optical connectivity is a good business. It obviously is.

The question is why the market treated the announcement of a $115 billion “category leader” deal like a warning sign instead of a reason to celebrate.

The logic behind this deal is not subtle, and neither company is shy about saying it out loud.

Lumentum’s CEO called the AI connectivity opportunity “enormous” in a CNBC interview the day the deal broke.

Marvell already sells custom AI silicon directly into hyperscaler datacenters.

Lumentum builds the optical transceivers and components that move data between GPUs, servers, and racks inside those same datacenters.

Right now, those are two separate purchase decisions, made by two separate teams, often from two separate vendors.

This deal collapses that into one vendor relationship, for one combined bill.

That’s the kind of positioning hyperscalers reward with long-term contracts, not quarterly purchase orders.

If AI datacenter buildout keeps compounding the way it has for three straight years, owning both ends of that pipe is a genuinely enviable position.

Here’s where the steelman case runs into a problem the market noticed immediately, even if most headlines didn’t.

The deal’s exchange ratio is fixed at 4.5 shares of Marvell stock for every share of Lumentum.

At Friday’s closing prices, before the announcement, that ratio implied a real, meaningful premium for Lumentum shareholders.

But an all-stock deal doesn’t lock in a dollar figure. It locks in a ratio.

The actual dollar value floats with whatever the acquirer’s stock does between announcement and closing.

Marvell’s stock didn’t just dip on the news. It fell nearly 13% in a single session, on triple its normal trading volume.

That drop dragged the implied value of the entire deal down with it, in real time, while the ink was still drying.

Run the math. At Monday’s close, 4.5 shares of a $171.41 stock works out to roughly $771 per Lumentum share.

Lumentum was trading near $841 before the deal was announced.

This isn’t a shrinking premium anymore. It’s a discount of close to 8%, on a transaction both companies pitched as a win for Lumentum’s owners.

Not a rounding error. A structural flaw in how this deal was priced from day one.

There’s a second question the bull case hasn’t answered.

Lumentum’s revenue leans heavily on a small number of hyperscaler customers buying optical gear for AI datacenters.

Marvell’s custom silicon business has nearly the identical problem. A handful of the same hyperscalers decide its growth rate every single quarter.

Combining the two companies didn’t diversify that exposure.

It doubled down on it.

Now one combined company’s fortunes depend entirely on the capital spending decisions of the same small customer list, twice over, with no offsetting business to cushion a slowdown.

And honestly? That’s the kind of concentration risk a buyer usually pays a premium to avoid, not one it pays $115 billion to deepen.

Marvell’s chart tells the story of a stock caught completely off guard by its own news.

Shares opened Monday near $192, gapped lower within the first hour of trading, and kept sliding until they briefly touched $168 before clawing back slightly into the close.

Marvell finished the day at $171.41, down 12.72% from Friday’s $196.40 close.

Volume told the real story here.

About 61.5 million shares changed hands, nearly three times Marvell’s typical daily volume, which is usually the fingerprint of forced or panicked selling, not a calm market repricing a deal.

The stock is now sitting roughly 48% below its 52-week high of $329.88, a level it hasn’t traded near since earlier this year, while still sitting well above its 52-week low near $70.69.

Lumentum’s chart looks even worse.

It fell as much as 18% intraday, the kind of single-day move you’d expect from a disappointing earnings report, not from a company that just agreed to be acquired at a “premium.”

If Marvell’s stock stabilizes anywhere near current levels, the deal math holds at roughly today’s discount, uncomfortable but survivable.

If it keeps drifting toward the low $160s or below, that discount widens further, and shareholders on both sides start asking louder questions about whether the boards actually locked in a fair deal, or just a fast one.

Four things will tell us whether this deal holds together, or whether Monday’s selloff was the first crack in it.

Watch for the actual S-4 filing, which will spell out the exact exchange ratio, any collar provisions, and whether Lumentum shareholders have any protection if Marvell’s stock keeps falling. If there’s no collar protecting the deal value, today’s discount can widen a lot further before the shareholder vote even happens.

A $115 billion combination inside the AI supply chain is exactly the kind of deal that invites a long antitrust look. Watch for the first regulatory comment period, and whether either company signals confidence or caution about how long closing will actually take.

I covered Marvell at a very different moment back in October, when the questions were about an earlier $90 billion bet. That piece is here. The next earnings print will show whether Marvell’s core chip business is strong enough to support two mega-bets running at once, or whether this deal is quietly papering over a slowdown in the business it already had.

Watch for any commentary from Lumentum’s biggest customers, officially or through analyst channels, on how they view the combination. If the companies that actually write the checks stay silent, that silence is itself a signal the deal has fewer committed allies than the press release implied.

I’ve sat with this one for a few days now, and here’s where I land.

Marvell didn’t make a bad strategic bet.

The logic behind owning both the silicon and the optics inside AI datacenters is sound, and probably where this business needed to go eventually.

What Marvell got wrong was timing and structure.

It locked in a fixed exchange ratio before the market had any chance to react, and the market reacted badly, fast, and with real volume behind it.

That alone doesn’t make this a bad deal. Plenty of good acquisitions trade down on announcement day before recovering.

But after October’s “math doesn’t add up” piece on Marvell’s last mega-bet, this isn’t the first time shareholders have been asked to trust the arithmetic on faith.

Markets don’t usually punish good ideas forever.

They do punish companies that ask for that kind of trust twice in one month.

This isn’t a reason to bail on Marvell.

It’s a reason to wait for the actual S-4 filing before deciding whether this “premium” deal is one at all.

At Cherry, we provide payment plans to 90% of pet parents. Plus, we make applying easy - with a 35-second online application they can submit right in the treatment room, and a soft credit check that doesn’t ding their credit history. Don’t let cost get in the way of care - with Cherry.

See how Cherry works

_{^{* These figures come from Cherry's provider case studies, so individual results vary. Details at withcherry.com/provider-reviews. Payment options through Cherry are issued by the following financing partners: withcherry.com/financing-partners}}

The real lesson from Monday has nothing to do with AI, or optics, or even Marvell specifically.

It’s about what “premium” actually means inside an all-stock deal.

A premium isn’t a fixed number written into a press release. It’s a bet that two stock prices will hold roughly steady long enough for the deal to close.

When the acquirer’s own stock drops 13% before the ink even dries, that premium was never real money sitting in anyone’s account.

It was a snapshot of a number that was always going to keep moving.

Every investor who’s ever celebrated a buyout premium on announcement day should sit with that for a second.

The real price of a deal isn’t set at the press release. It’s set on the day it actually closes, and not a day before.