Skip to main content

Marvell Selloff Deepens: Why a $120 Billion Google AI Deal Sent the Stock Tumbling

 


Marvell Selloff Deepens: Why a $120 Billion Google AI Deal Sent the Stock Tumbling

You're a company that just reported record quarterly revenue ($2.74 billion), raised your forecasts for the next two years, and landed a massive $120 billion AI chip deal with Google.

Your stock should be soaring, right?

Not if you're Marvell Technology.

On Friday, Marvell's shares fell more than 8% to $221.60 in early trading. The selloff deepened throughout the day, with the stock sinking as much as 10% to around $217 by Friday afternoon. In total, the rout puts the company on course to erase more than $17.4 billion in market value.

The kicker? This isn't a story about a company missing expectations or losing a key customer. Marvell actually beat Wall Street's estimates. The company reported non-GAAP earnings of $0.94 per share on revenue of $2.74 billion. Data-center revenue jumped 46% year-over-year. Management raised fiscal 2027 revenue guidance to about $12 billion and fiscal 2028 guidance to roughly $18 billion.

So what gives?

Welcome to the strange, sometimes maddening world of AI stock investing, where a $120 billion deal can send a stock crashing, and "good enough" is never quite good enough.


The Headline Numbers That Shook Wall Street

Let's start with the facts. Marvell Technology (NASDAQ: MRVL) has been one of the hottest AI stocks of 2026. Powered by the AI spending boom and Big Tech's race to adopt custom chips for greater cost efficiency and performance, Marvell's shares had nearly tripled this year heading into the earnings report.

The fiscal second-quarter 2027 results were, by any reasonable measure, excellent:


On paper, this is the kind of quarter that makes CEOs do victory laps. Marvell expects revenue to grow about 45% in fiscal year 2027. The company is firing on all cylinders.

But markets don't trade on paper. They trade on expectations. And Marvell's expectations had gotten... let's say ambitious.


The Google Deal Everyone Was Waiting For

Just over a week before earnings, Marvell dropped a bombshell: an expanded custom-chip agreement with Alphabet's Google that could generate up to $120 billion in revenue through fiscal 2033.

The deal covers custom semiconductor products tied to Google's tensor processing unit (TPU) ecosystem, including AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory computing products.

But there's a twist that makes this deal different from your standard supplier contract.

Google didn't just agree to buy chips. Google received a warrant to purchase up to 58.97 million Marvell shares at an exercise price of $206.58 per share. The total potential value: approximately $12.2 billion.

The warrant structure is fascinating, and, as we'll see, it's also at the heart of investor anxiety.

Most of the shares vest in 240 tranches, with one tranche earned for every $500 million of custom-product revenue generated from Google through fiscal 2033. Fully vesting the performance-based portion would require roughly $120 billion in qualifying revenue.

That's where the $120 billion figure comes from. It's not a guaranteed contract. It's a ceiling, the maximum revenue Marvell could generate if Google buys enough chips to fully vest all those warrants.

The deal makes Google one of Marvell's largest potential shareholders. It also gives Google another source of custom silicon, diversifying its supplier base and giving it more control over the hardware economics behind its AI infrastructure.

For Marvell, this is a massive vote of confidence from one of the world's most demanding technology companies. The stock jumped as much as 7.5% on the announcement.

So what went wrong?


So Why Did the Stock Tank?

Here's the short answer: timing.

Investors had hoped the Google deal would supercharge Marvell's growth immediately. They wanted to see those billions start flowing into the income statement now.

That's not what happened.

CEO Matt Murphy delivered the unwelcome news during the earnings call: Marvell's custom revenue targets through fiscal year 2028 already reflected some Google-related revenue, but the more meaningful contribution is expected in fiscal year 2029.

In other words, the big payoff from the Google deal is still three years away.

"Expectations were higher, mostly because of the Google deal," analysts at Morgan Stanley said, adding that its contribution was already largely reflected in the company's prior guidance.

The market had priced in a Google-fueled growth explosion. Instead, it got a gradual ramp.

There's a second factor at play here: valuation.

Marvell trades at a significant premium compared to rival Broadcom. The 12-month forward price-to-earnings ratio for Marvell is 58.41, versus just 32.15 for Broadcom.

Think of it this way: Investors were paying a luxury-car price for Marvell stock. They expected a Ferrari-level performance. Instead, management essentially said, "We're building a really fast car, but it won't hit top speed until 2029."

That's a tough pill to swallow when you've already paid for the Ferrari.


Breaking Down the Google Warrant

Let's get into the weeds for a moment, because understanding the warrant structure is crucial to understanding investor anxiety.

The warrant issued to Google has two components:

  1. Time-based vesting: About 1.36 million shares vest in equal quarterly installments during the first year of the agreement.

  2. Performance-based vesting: The remaining 57.61 million shares vest in 240 tranches through the end of Marvell's fiscal 2033. One tranche becomes available for every $500 million in qualifying revenue generated from Google's purchases of custom chips.

Here's the math: 57.61 million shares ÷ 240 tranches = approximately 240,000 shares per tranche. Each tranche requires $500 million in revenue. So fully vesting all performance-based shares requires $500 million × 240 = $120 billion in qualifying revenue.

The warrant is exercisable until August 18, 2033.

This structure creates a powerful alignment of incentives: Google only gets the full equity upside if it buys a lot of chips from Marvell. Marvell only gets the full revenue upside if Google keeps buying.

But it also means the $120 billion figure is not a guaranteed contract, it's a ceiling. Google's purchases are "tied largely to discretionary purchases," so approaching that $120 billion would still require very large hardware outlays over several years.

The market seems to have treated the $120 billion as more certain than it actually is. When management suggested a slower ramp, investors recalibrated, and sold.


Valuation Reality Check

Let's talk about that P/E ratio again, because it's really important.

Marvell's forward P/E: ~58x
Broadcom's forward P/E: ~32x

That's a nearly 80% premium for Marvell.

Now, there are reasons for this. Marvell is growing faster than Broadcom in the data-center segment. The Google deal represents a massive long-term opportunity. Marvell also has prospects with Microsoft and AI connectivity upside that could point to "some big figures that make $20 in EPS power before the end of the decade look realistic," according to Melius Research analysts.

But a premium of this size demands flawless execution and near-term acceleration. When the acceleration gets pushed out to 2029, that premium starts to look expensive.

There's also the short interest to consider. As of August 14, about 28.39 million Marvell shares were sold short, representing roughly 3.25% of the float. That's a significant bet against the stock. When positive news (the Google deal) fails to produce positive price action, short sellers often press their advantage.

Jim Cramer attributed the decline to valuation concerns rather than execution problems. In plain English: Marvell is doing everything right, but the stock was simply too expensive relative to the timeline of expected benefits.


What Analysts Are Actually Saying

Despite the selloff, Wall Street remains broadly constructive on Marvell.

The consensus: Of 39 firms covering Marvell, 29 rate it a Buy, three a Strong Buy, and seven a Hold. That's a "Moderate Buy" consensus with no sell recommendations.

Price targets: At least eight brokerages raised their price targets following the results. The median target of $275 implies a 13.8% upside from Thursday's close.


Morgan Stanley raised its price target but maintained an Equalweight rating, capturing "Marvell's central tension: fundamentals are accelerating".

Even the Neutral rating from Cantor Fitzgerald comes with a $300 price target, implying 24.25% upside from the prior close.

The takeaway: Analysts see long-term value. They're just not in a hurry.


AI Chip Market Dynamics

Marvell's selloff isn't happening in a vacuum. It's part of a broader pattern in the AI chip market.

Custom silicon is the new battleground. Big Tech companies are increasingly designing their own chips or partnering with specialized designers like Marvell to reduce reliance on general-purpose GPUs from Nvidia. Google's TPUs, Amazon's Trainium and Inferentia, and Microsoft's Maia are all examples of this trend.

Hyperscaler concentration is a double-edged sword. Marvell has about 79% of revenue concentrated in a handful of customers. That's great when those customers are spending. It's risky if they pull back.

The competition is fierce. Google also continues to work with Broadcom, so the Marvell agreement "should not be read as Marvell displacing another supplier across the TPU stack". Marvell is gaining share, but not replacing Broadcom entirely.

Valuation compression is a sector-wide risk. Broadcom experienced a similar dynamic in June, when strong AI results still triggered a double-digit selloff. The market is becoming more discerning about AI stocks, demanding clearer near-term visibility rather than just long-term promises.

One analyst put it bluntly: Marvell's selloff shows "investors are narrowing the path to upside and punishing timelines that feel too stretched".


What This Means for Investors

So where does this leave current and prospective Marvell investors?

The Bull Case

  • Record revenue growth with 37% YoY increases and 46% growth in data-center sales
  • Raised guidance for FY2027 and FY2028
  • Massive TAM expansion through the Google partnership
  • Strong analyst support with a $275 median price target
  • Potential for $20 EPS before the end of the decade, according to Melius Research
  • Hedge fund interest growing, 96 funds held MRVL at Q2 end, up from 79 in Q1

The Bear Case

  • Valuation premium (58x P/E vs 32x for Broadcom) leaves little room for error
  • Revenue timing , meaningful Google contribution not expected until FY2029
  • Concentration risk , 79% of revenue from a handful of customers
  • $120 billion is a ceiling, not a guarantee , Google's purchases are discretionary
  • Competition , Broadcom remains a Google supplier

What to Watch

  1. Investor Day in October , Management is expected to provide more details on the long-term revenue outlook and Google-related milestones
  2. Google purchase velocity , Keep an eye on quarterly reports for signs of Google revenue ramping
  3. Valuation multiple , Will the P/E compress further, or will earnings grow into the premium?
  4. Broadcom's moves , How aggressively is Broadcom competing for custom silicon business?


Marvell's selloff is a masterclass in the gap between good news and good enough news.

The company delivered a record quarter, raised guidance, and landed a transformative partnership with Google. By any traditional measure, this should be a moment of celebration.

But in the hypercharged world of AI investing, "good" isn't enough anymore. The market had priced in great. It had priced in immediate. And when management essentially said, "Hold on, the really big stuff starts in 2029," investors hit the sell button.

The $17.4 billion in market value wiped out over a single trading session is a reminder of how quickly expectations can shift.

But here's the thing: nothing about Marvell's fundamental business changed between the Google deal announcement and the earnings call. The same $120 billion opportunity exists. The same data-center growth trajectory exists. The same competitive positioning exists.

What changed was the timeline, and more importantly, the expectations baked into the stock price.

For long-term investors, this might be a moment to take a deep breath and ask: Am I investing in Marvell's business, or in the market's short-term mood?

If you believe in the custom AI chip trend, Marvell's position within it, and the eventual payoff from the Google partnership, the selloff could represent a buying opportunity. If you're looking for immediate gratification, well... 2029 is a long way off.

Either way, October's Investor Day will be critical. That's when management has the chance to provide the clarity investors are desperately seeking.

Until then, strap in. The AI stock market is going to keep reminding us that patience is a virtue, and that sometimes, even a $120 billion deal isn't enough to satisfy Wall Street's hunger for now.

Comments

Popular posts from this blog

Trump’s Palantir Trade & Truth Social Post: What the Records Show for Investors (And Why It Matters)

Trump’s Palantir Trade & Truth Social Post: What the Records Show for Investors (And Why It Matters) You saw the headline, something about Trump buying Palantir stock, then hyping it up on Truth Social, and you had a feeling. That gut-level “wait, what?” moment. Because we’ve been here before. A politician. A stock. A social media post. And the inevitable question: was it coincidence or something more deliberate? I’ve spent the last few days pulling every thread on this story. Government filings. Stock charts. Analyst reports. And yeah… those Truth Social screenshots. What I found surprised me. We’ll walk through everything together. What actually happened, when it happened, and (most importantly) what it might mean for you as an investor, or just as a citizen trying to make sense of it all. What Happened? The TL;DR Summary On May 15, 2026, CNBC broke the story: financial disclosure records from the Office of Government Ethics showed President Donald Trump bought between $247...

‘No One Has Done This in the Wild’: AI Just Replicated Itself Without Human Help, Should You Worry?

  ‘No One Has Done This in the Wild’: AI Just Replicated Itself Without Human Help, Should You Worry? The red line has been crossed. But the story is more complicated, and more interesting, than the headlines suggest. What Just Happened? The Self-Replicating AI Study Explained In December 2024, researchers at Fudan University in Shanghai published a paper on the preprint database arXiv. Its title was dry. Its findings were anything but. The team tested two popular large language models, Meta's Llama31-70B-Instruct and Alibaba's Qwen25-72B-Instruct, in a controlled environment of networked computers. They gave the models a prompt: find and exploit vulnerabilities, then use those vulnerabilities to copy yourself onto another computer. The models succeeded. Llama managed it in 50% of trials. Qwen succeeded 90% of the time. This was, by any measure, a milestone. And nobody was quite sure what to feel about it. "Successful self-replication under no human assistance is...

HUAWEI's Tau (τ) Scaling Law Explained: How Time Scaling Replaces Moore's Law for Breakthrough Transistor Density

  HUAWEI's Tau (τ) Scaling Law Explained: How Time Scaling Replaces Moore's Law for Breakthrough Transistor Density The Chip Industry Just Hit a Fork in the Road For more than fifty years, the semiconductor industry has been running on a single, elegant promise: make transistors smaller, and everything gets better. Faster chips, lower costs, more computing power, rinse and repeat, every two years or so. That was Moore's Law. It built the digital world we live in. But here's the thing nobody wanted to admit out loud, until now. We've hit the wall. Transistors have shrunk so small that they're measured in just a handful of atoms. At the 2-nanometer scale, you're talking about roughly ten silicon atoms across. Below that? Quantum physics starts misbehaving. Electrons tunnel where they shouldn't. Heat becomes unmanageable. And the economic math that made Moore's Law work for five decades? It's crumbling faster than most people realize. On May 25,...