To The Moon

Wednesday's $NVIDIA(NVDA)$ and Thursday's $Marvell Technology(MRVL)$ : Two Strong Earnings Reports, Two Completely Different Scripts

$MRVL Marvell plunged after hours, immediately erasing the gains of the past few days.

After reviewing $NVDA NVIDIA's and Marvell's earnings back-to-back, my biggest takeaway is this: both delivered excellent results, yet the market applied two entirely different pricing frameworks.

• NVIDIA: sold off first, then rallied

  

• Marvell: strong numbers, stock got hammered

  

Explaining this away as simple "buy the rumor, sell the news" feels insufficient. The real difference lies in how much the market had already priced in before the reports were released.

First, let's look at $NVIDIA(NVDA)$

Q2 revenue came in at $96.2 billion, up 106% year-over-year. Data center revenue reached $89.0 billion, up 117%. For the next quarter, the company guided to $108.0 billion ±2%.

These figures are already staggering. Yet when the report first dropped, the after-hours reaction was lukewarm—the stock even dipped briefly.

The reason is straightforward. For NVIDIA today, merely beating Wall Street expectations is no longer enough. The market has grown accustomed to eight consecutive quarters of beats. Ahead of the report, the options market was pricing in only about a 5.4% implied move. Everyone simply assumed another strong quarter.

So when the numbers first hit, here's what traders saw:

• Revenue: great.

• Data center: great.

• Guidance: great.

But all of it was more or less within market expectations.

What truly changed the entire trading structure came during the earnings call—and specifically, during the Q&A with investors.

Analysts on the call pressed management on the sustainability of AI demand and the trajectory beyond the current cycle. Jensen Huang didn't just defend the outlook; he went on the offensive. He emphasized that AI demand is not plateauing but accelerating, and dropped a bombshell: Rubin has already entered full-scale production. More critically, when pushed on the long-term growth trajectory, management guided that FY2028 revenue could still grow by approximately 70%.

You could almost feel the tone shift in real time. The buy-side analysts who had dialed in looking for reasons to trim their positions suddenly found themselves recalculating models. The question that had haunted the market—if NVIDIA is already doing nearly $100 billion per quarter, how much bigger can this get?—was answered with a number larger than most had penciled in.

This created a beautiful sequence for NVIDIA:

Earnings release → market feels underwhelmed → stock dips → investors grill management on the call → management delivers a stronger-than-expected long-term vision → short covering → longs chase back in.

The next day, NVIDIA finished up 8.7%.

Now, $Marvell Technology(MRVL)$

On the surface, it's genuinely hard to find major flaws in its report.

Q2 revenue: $2.739 billion, up 37% YoY, another all-time high. Non-GAAP EPS hit $0.94. Data center grew 46% YoY. Q3 revenue guidance was raised further to $3.15 billion ±5%.

More importantly, the company raised its long-term targets again: FY2027 revenue was lifted from roughly $11.5 billion to $12 billion; FY2028 was raised from $16.5 billion to $18 billion.

By any financial metric, this is a solid earnings report.

Yet Marvell dropped over 6% after hours.

Why?

Because the market had already traded a much sexier story before the report. That story's name was Google.

A week prior, Marvell announced an expanded custom silicon partnership with Google. Under the agreement, Google could potentially bring up to ~$120 billion in procurement to Marvell over time, while gaining up to ~$12.2 billion in equity exposure to Marvell.

After that news, the market easily formed a mental image: Marvell's original growth curve was one trajectory; with Google entering, that curve should suddenly steepen.

So what the market was really waiting for was: how much further could Google push future revenues?

The tension peaked during the Q&A session. Investors weren't asking about Q2—they were asking about the Google partnership and whether it would force a wholesale upward revision to the FY2028 target. Analysts pressed CEO Matt Murphy for specifics: Is the $18 billion number conservative? Does it fully capture the Google opportunity? Should we model something materially higher?

Murphy's answers cooled those expectations. He made it clear that Marvell's existing FY2028 custom silicon revenue target already includes some contribution from Google. The more material Google revenue contribution, he said, wouldn't really show up until FY2029.

And regarding the previously stated FY2029 custom silicon target of "above $10 billion," while the company acknowledged upside potential, they did not immediately offer a new, higher number—instead deferring further details to the October 6 Investor Day.

You could sense the disappointment through the screen. The buy side had dialed in hoping to hear that the Google deal was an additive layer on top of the existing $18 billion framework. Instead, they learned it was already baked into that framework.

So a story that had been projected into the distant, expansive future was pulled back onto a more realistic timeline.

Marvell's fundamentals didn't suddenly deteriorate. AI orders remain robust. Data center is still growing 46%. Custom silicon is expected to double next year. FY2027 and FY2028 revenue targets were both raised.

It's just that the stock had already front-run far more of the future.

Heading into earnings, MRVL had nearly tripled year-to-date. Once a company is being priced on the narrative of "Google + Custom ASIC + AI infrastructure's second growth curve," the evaluation criteria for its earnings report naturally shift.

Putting the two reports side by side makes for a fascinating contrast.

Ahead of NVIDIA's report, the market's question was: Can AI growth continue?

During the investor Q&A, the call provided an answer stronger than the market had imagined. So the stock was re-priced upward.

Ahead of Marvell's report, the market's question had already become: How fast can Google make Marvell grow?

During the investor Q&A, the call confirmed the growth was real, but that large-scale realization would take more time. So the market began compressing near-term valuation.

This also explains why two companies with strong earnings produced completely opposite candlesticks.

• NVDA followed the script of: expectations were held down, then the investor Q&A raised them.

• MRVL followed the script of: expectations were already maxed out, then the investor Q&A recalibrated the timeline.

In earnings trading, a seemingly counterintuitive phenomenon often appears:

Great results, stock drops anyway. Slightly imperfect results, stock rockets.

Because what a stock always trades on is the gap between reality and what the market had previously imagined that reality to be.

From this perspective, I wouldn't write off Marvell simply because of last night's candle.

If what follows is continued upward earnings revisions, with the stock merely digesting valuation after prior expectations ran too hot, then this post-earnings drop looks more like a timeline repricing.

If, however, the pace of revenue realization consistently falls short of the valuation the market had previously assigned, that is when the underlying thesis truly needs to be re-examined.

What truly matters during earnings season was never just the beats or misses.

It's also how much of the future the market had already believed in before the report—and how the investor Q&A either confirmed or challenged that belief.