The Moment That Changed the Story

Forget the earnings calendar. The catalyst that just sent Marvell (NASDAQ: MRVL) ripping more than 15% in overnight trading wasn't a number on a 10-Q. It was a single sentence, delivered live by the most powerful CEO in tech.

On June 2 in Taipei, Jensen Huang made a surprise appearance during Marvell CEO Matt Murphy's Computex 2026 keynote. He stayed on stage for roughly ten minutes. Before walking off, he turned to the audience and said:

"This is the next trillion-dollar company, ladies and gentlemen."

That isn't a tweet from an anonymous fund manager. That's the CEO of the world's most valuable company — sitting on a $5T+ market cap — publicly anointing Marvell as the next member of the four-comma club, on the keynote stage of the biggest hardware event of the year. The market did exactly what it should have done: it re-rated the stock on the spot.

The Tape Was Already Telling You Something

This wasn't a stock that needed saving. MRVL was already one of the most powerful charts in semis going into Computex. The stock is up roughly 130% year-to-date in 2026, and an even more astonishing 175% since early February. It closed June 1 at around $223, already up ~9% on the day from the Teralynx T100 announcement — and then the overnight tape added another 15%+ on Jensen's comment, pushing market cap to roughly $199 billion. That still leaves about 5x of upside to hit the trillion-dollar target Huang just laid out.

When a stock this strong gets a catalyst this clean, it doesn't just gap and fade. It re-rates.

What Marvell Actually Sells (And Why It Matters Now)

A lot of investors still file Marvell under "old networking chip company." That mental model is roughly two years out of date.

Marvell has quietly rebuilt itself into one of the two or three most critical names in AI infrastructure outside of NVIDIA itself. Per the company's most recent fiscal disclosures, data center is now ~76% of revenue, up from 50% just two years ago. Fiscal Q1 came in at a record $2.418 billion in revenue, +28% year-over-year, with record operating cash flow of $638.8 million and non-GAAP EPS of $0.80.

The product portfolio sits on top of three of the biggest chokepoints in the AI build-out. The first is custom silicon — the XPUs and ASICs that power Amazon's Trainium and similar programs at Google, Microsoft, and Meta. This is the business going head-to-head with Broadcom, and the one Wall Street is willing to assign the highest multiple to. The second is optical DSPs and silicon photonics: every photon moving between GPUs in a modern AI cluster passes through a DSP, and Marvell owns this market — a position reinforced by the recent Celestial AI acquisition, which doubles down on silicon photonics, the technology that lets bandwidth keep scaling without melting power budgets. The third is AI-era Ethernet switching: on June 1, Marvell announced the Teralynx T100, which the company describes as the industry's first 102.4 Tbps single-chip switch, built on 3nm and claiming up to 25% lower power than competing solutions. This is the chip hyperscalers have been waiting for to flatten their AI fabrics and stop burning power on extra switch tiers.

In one sentence: NVIDIA sells the GPUs. Marvell sells almost everything that connects them.

Why Jensen's Trillion-Dollar Call Is More Than Hype

It would be easy to dismiss the Computex moment as showmanship. It isn't.

Start with the fact that Jensen is putting NVIDIA's money where his mouth is. In March 2026, NVIDIA disclosed a $2 billion direct investment in Marvell and brought it into the NVLink Fusion ecosystem. Around the same window, NVIDIA also wrote $2B checks to Lumentum and Coherent. Jensen isn't speculating about the optical interconnect layer — he's quietly buying it.

Then there's the deeper structural shift. The AI bottleneck has officially moved from compute to connectivity. Rack power is pushing toward 120kW, GPUs scale faster than the networks around them, and hyperscalers are no longer compute-constrained at the chip level — they're constrained by latency, bandwidth, and the power their optical links burn. Murphy said it plainly from the Computex stage: connectivity is the next AI bottleneck, and Marvell is the "undisputed connectivity leader." Jensen stood next to him and agreed.

The platform has the depth to back the narrative. Per Murphy's own keynote remarks, Marvell has invested roughly $36 billion rebuilding its data infrastructure platform. It skipped 7nm and 14/16nm entirely, jumped straight to 5nm, and is now shipping 3nm in lockstep with TSMC's leading edge. With custom XPUs, optical DSPs, and Ethernet switching all under one roof, Marvell can monetize the same AI cluster three different ways. That's the kind of structural setup that supports a trillion-dollar valuation — not a one-quarter momentum trade.

The Bull Case Going Forward

The setup tonight is about as clean as it gets. AI capex is accelerating, not peaking — hyperscaler 2026 budgets are tracking toward $650B+, and the spending is rotating toward the layers Marvell owns. Custom silicon is the highest-margin growth vector in semis, and Marvell is the cleanest pure-play challenger to Broadcom. The T100 doesn't ramp meaningfully until 2027, which means Jensen's "trillion-dollar" call is being made before the new product cycle even hits the income statement. Layer on NVIDIA's $2B equity stake — a strategic floor that simply did not exist six months ago — and the fact that 76% of revenue is now AI-leveraged, and Marvell has effectively transformed itself into an AI infrastructure company that happens to wear a legacy ticker.

For the trillion-dollar target to play out, MRVL needs to roughly 5x from here. Aggressive? Yes. Impossible? Ask anyone who said the same thing about NVIDIA at $300 billion. The path runs through one specific equation: custom silicon revenue scales the way bulls think it will, and the optical layer remains a Marvell monopoly. Hit both, and the math gets there.

Bottom Line

Tonight wasn't just a 15% pop. It was the moment Marvell got publicly reclassified — by the single most influential voice in technology — from "AI beneficiary" to "AI infrastructure platform with a trillion-dollar runway."