Marvell’s Moat: Why Jensen Huang Called It the Next $1 Trillion Company
At Computex on June 2, Nvidia’s Jensen Huang stood next to Marvell’s CEO and called it “the next trillion-dollar company” — because the AI data center is useless without the connectivity and optics layer Marvell owns. The stock jumped about a third in a day. The June 7 data says the moat is real — revenue is set to roughly double to ~$17B in two years — but the price already borrowed years of it: Marvell trades at 65× forward earnings (about 3× Nvidia), ~32% above every published analyst target, and insiders sold a net $107M into the run. This is the moat, the math behind the trillion-dollar call, and what has to be true to get there.
Marvell’s 20-day run around Jensen Huang’s $1 trillion call
+55%
▲ +54.86 a one-month surge after Nvidia’s CEO called Marvell the next trillion-dollar company at Computex — the stock now trades at 65× forward earnings and above every published analyst target.
Jensen Huang does not hand out trillion-dollar labels. On June 2 in Taipei he gave one to Marvell — “the next trillion-dollar company” — and the stock ripped ~32% in a session. His logic is the most defensible part of the bull case: as AI is broken into thousands of chips spread across a data center, the bottleneck stops being compute and becomes connectivity — the interconnect and optics that move data between GPUs — and that is the layer Marvell owns, alongside a custom-silicon business it builds for the cloud giants. The growth is real: revenue is modeled to roughly double from ~$8B to ~$17B in two years, +40–45% a year, with estimates raised 25-to-3. The catch is the price. A $1 trillion Marvell is roughly 5× from here and would demand a multiple richer than Nvidia has ever held. Today the stock already trades at 65× forward earnings, ~32% above the Street’s highest published target — and the people who run it sold a net $107M into the surge. The moat is real; the question is how many years of it you are paying for on day one.
The contradiction
The moat is real: Marvell owns the connectivity and custom-silicon layer Jensen says AI cannot scale without, and revenue is set to roughly double in two years. But “next trillion-dollar company” is a ~5× re-rating the price already started to front-run — the stock trades above every published target, at 65× earnings, and the insiders sold into the pump.
What the headline says
What Jensen Huang told Computex
Marvell is the next $1 trillion company
What the data says
What the June 7 data says
~5× away, 65× earnings, above every target — and insiders sold a net $107M
Chapter 01
The anointment: Jensen Huang’s $1 trillion call, and the surge
On June 2 at Computex, Jensen Huang called Marvell “the next trillion-dollar company” while standing next to its CEO. The stock jumped ~32% that day and still closed the week up 20% — even after a 17% Friday — while the rest of the AI-infrastructure complex was red.
Start with what actually happened, because the catalyst here is unusually clean. On Monday, June 2, at Computex in Taipei, Nvidia CEO Jensen Huang shared a stage with Marvell CEO Matt Murphy and said the quiet part out loud: Marvell, he told the room, is “the next trillion-dollar company.” Coming from the man whose own company is the most valuable on earth and who has spent two years narrating the AI build-out, that is not a throwaway line — it is the single most valuable endorsement in the industry, and the tape reacted instantly. Marvell ripped roughly 32% that session. By the close of the week it had given a chunk back — down about 17% on Friday alone — and still finished up 20% on the week, while almost everything else in the AI-infrastructure complex was bleeding: Broadcom fell 16% on the week, Nvidia and Arista each fell around 9%, Credo 8.5%. Over 20 trading days Marvell is up nearly 55%. Huang’s logic, which the next section unpacks, is the most defensible part of the bull case. The rest of this piece is about the gap between that logic and the price it has already produced.
"The next trillion-dollar company, ladies and gentlemen." — Jensen Huang, Computex, June 2, 2026
Marvell vs the AI-infrastructure complex: 5-day return, week of June 1–5
Source: MarketDecode weekly scanner snapshot, June 5 close
Week of June 1–5
20.1%green while the complex bled
The day (June 2)
~32%Huang’s Computex remark (tier-1)
20-day run
54.9%the move the price has to defend
Chapter 02
The moat: the connectivity layer AI can’t skip
Huang’s argument is the moat: as AI is split across thousands of chips, the bottleneck becomes connectivity — the interconnect and optics that move data between GPUs — and Marvell owns that layer plus a custom-silicon business for the cloud giants. Revenue is modeled to roughly double from ~$8B to ~$17B in two years.
Here is why Huang said it, in plain terms. Modern AI does not run on one chip; it runs on thousands of them, wired together across an entire data center, and the model only works if those chips can share data fast enough to act like one machine. “When you take a computing problem and disaggregate it into a lot of parts and distribute it across the data center, what’s necessary is connectivity,” Huang said — “that’s the reason why Marvell is so essential.” That is the moat: Marvell sells the interconnect and electro-optics that move bits between GPUs at the speed the cluster demands, and it co-designs custom AI silicon for the hyperscalers — the in-house chips cloud giants like Amazon and Microsoft use to avoid buying every accelerator from Nvidia. Nvidia put money behind the words, taking a reported $2 billion stake and partnering with Marvell on its NVLink ecosystem. And the financials back the story: consensus has Marvell’s revenue growing about 41% this fiscal year and 45% next — from roughly $8.2 billion to about $16.7 billion in two years — with EPS rising toward $6.17, and estimate revisions running 25 up to just 3 down. This is a real, accelerating business in the single most important spend cycle in tech. That is the half of the story that earns the headline. The other half is what you pay for it.
"That’s the reason why Marvell is so essential." — Jensen Huang on the connectivity bottleneck
Marvell revenue trajectory: on track to roughly double in two years
Source: MarketDecode forward estimates, June 7, 2026 (consensus revenue; prior year derived from YoY)
Revenue growth
41% → +45%this FY to next FY, accelerating
EPS next FY
$6.17+53% YoY on consensus
Estimate revisions
25 up / 3 downthe Street is raising into it
Chapter 03
What you pay: 65× earnings — and the trillion-dollar math
Marvell trades at 65× forward earnings — roughly 3× Nvidia (23×) and 2× Broadcom (33×). And “next trillion-dollar company” is not a price target: from ~$190B it implies ~5×, and a price-to-sales multiple near 40 — richer than Nvidia (~21×) has ever sustained.
Now the price. At 65× forward earnings, Marvell is the most expensive name among the proven AI-chip leaders — about three times Nvidia’s 23× and twice Broadcom’s 33×, the very company that competes with it most directly in custom silicon. The only names trading richer are smaller, earlier-stage optics and connectivity plays — Coherent at 69×, Lumentum and Astera Labs above 100× — which we cite but do not chart, because they would flatten the comparison that matters: against the giants with real earnings, Marvell already carries the premium multiple. Then there is the headline itself. “The next trillion-dollar company” sounds like a target; it is really a multiplier. Marvell was worth roughly $190 billion when Huang said it, so a $1 trillion Marvell is about five times bigger than today. Run the math the way the analysts at the Motley Fool did and it gets concrete: to reach $1 trillion on management’s own guidance, Marvell would need a trailing price-to-sales ratio near 40 by 2029 — or a price-to-earnings ratio around 127 on current estimates. For reference, Nvidia — the most important company in the build-out — trades at about 21 times sales and 33 times earnings, and never sustained anything close to 40× sales at this scale. So the trillion-dollar call is not wrong; it is a bet that Marvell out-executes its own guidance for years and the market keeps paying a multiple no megacap chipmaker has held. The moat can be real and the number can still be a stretch. Both are true at once.
"The moat can be real and the number can still be a stretch. Both are true at once."
Forward P/E: Marvell vs the proven AI-chip leaders
Source: MarketDecode valuation snapshot, June 7, 2026 (forward P/E)
Marvell
65×priciest of the profitable leaders
The giants
NVDA 23× · AVGO 33×cheaper, larger, proven
To $1 trillion
~5× from hereneeds P/S ~40 (NVDA ~21)
Chapter 04
The engine, decoded: a momentum stock, not yet a quality one
Decode the factor scores and the run explains itself. Marvell’s profile spikes on momentum (77) but trails Nvidia on value (49 vs 72), quality (48 vs 65) and opportunity (43 vs 74). The engine sees the better trade, not yet the better business.
Strip the narrative away and look at what the factor engine sees, because the shape of it is the whole tell. We score every name on momentum, value, quality and opportunity from 0 to 100. Marvell’s fingerprint is lopsided: momentum a towering 77 — the highest in the group, which is exactly what a 55% run produces — but value just 49, quality 48, and opportunity (the engine’s read on risk-adjusted upside from here) only 43. Lay Nvidia’s profile on top and the contrast is stark: Nvidia scores lower on momentum (52, because it has been flat-to-down) but far higher everywhere that describes the underlying business — value 72, quality 65, opportunity 74. In plain English, the model is telling you Marvell has been the better trade and Nvidia is still the better business. That is not a knock on Marvell’s moat; it is a statement about price. A stock that is all momentum and middling on quality, value and forward opportunity is a stock whose recent return came from the story re-rating, not from the fundamentals catching up — yet. For the radar to fill out, the next several quarters of the custom-silicon ramp have to convert that 77 of momentum into quality and opportunity. Until they do, you are buying the polygon’s sharp corner, not its area.
"Marvell has been the better trade. Nvidia is still the better business."
Factor fingerprint: Marvell vs Nvidia
Source: MarketDecode scanner factor scores, June 7, 2026 (0–100)
Momentum
MRVL 77 vs NVDA 52the 55% run, quantified
Quality
MRVL 48 vs NVDA 65the business gap
Opportunity
MRVL 43 vs NVDA 74forward upside the engine sees
Chapter 05
The analyst tape: the price ran past every target
At $263, Marvell trades above every published analyst target — RBC $200, UBS $195, TD Cowen $190 (Hold), Evercore $155, Benchmark $130. The stock sits ~32% above the Street’s highest mark, and every one of those targets is dated before Jensen spoke.
Wall Street has not caught up to Jensen Huang — and that is the cleanest evidence the price ran ahead of the models. Look at the most recent published price targets and they cluster in a band the stock has already left behind: RBC Capital at $200, UBS at $195, TD Cowen at $190 with a Hold, Evercore at $155, Benchmark at $130. Marvell closed at $263.47. That means the stock trades roughly 32% above the highest target any of these desks had on it — and below none of them. Crucially, every one of those targets is dated before June 2: the Street’s numbers were set when Marvell was a strong custom-silicon and optics story, not a trillion-dollar candidate. So the price you see today is not the Street’s view; it is Huang’s view, front-run by the market before a single analyst rebuilt a model around it. There are only two ways that gap closes. Either analysts raise their targets to chase the price — the bull path, and the thing to watch for in the coming weeks — or the price falls back toward the targets. A stock priced above every published number is not necessarily wrong, but it has spent its margin of safety: from here, the burden of proof is entirely on Marvell to make the analysts move, rather than on the analysts to justify the stock.
"The price you see is not the Street’s view. It’s Huang’s view, front-run."
Recent Wall Street targets for Marvell vs today’s price
Source: MarketDecode scanner snapshot, June 7, 2026 (analyst.recentActions; all dated pre-June 2)
Street high
RBC $200−24% below the price
Today
$263above every published target
The catch
all pre-June 2set before Huang spoke
Chapter 06
Who’s selling — and the verdict
Into the anointment, Marvell insiders sold a net $107M (36 buys vs 80 sells) — the heaviest in the connectivity complex, where Nvidia (−$98M) and Broadcom (−$79M) also sold and only TSMC bought. The verdict: real moat, borrowed price. Watch the 20-day line near $208 and the December 1 print.
Close on the people who know the company best. Over the last 90 days — the window that contains the run — Marvell insiders were net sellers to the tune of $107 million, 36 buys against 80 sells, the heaviest net selling in the connectivity complex. They were not alone: Nvidia insiders sold a net $98 million on zero buys, Broadcom $79 million on zero buys, Arista and Lumentum smaller amounts. The lone insider buyer in the whole group sits at the foundry — Taiwan Semiconductor, 60 buys against a single sell. Insider sales are never a timing signal on their own; executives sell for taxes, diversification, a hundred reasons. But as a body of evidence stacked on a 65× multiple, a price above every target, and a factor profile that is all momentum, the insider tape leans the same way the rest of the data does: the run got ahead of the people running the company. So the verdict. This is not “short Marvell.” The moat is real, Huang’s logic is sound, and the growth is genuine — Marvell may well be a far bigger company in five years. The thesis is narrower: the price has already borrowed years of that future, and the margin of safety is gone. The graders are concrete. Watch whether analysts raise targets toward the price (the bull confirmation), whether Marvell holds its 20-day line near $208 (lose it and the Jensen pump unwinds), and the December 1 earnings print — the first hard read on whether the custom-silicon ramp is converting the trillion-dollar story into trillion-dollar numbers.
"The price has already borrowed years of the future. The margin of safety is gone."
Insider net buying across the connectivity complex, last 90 days
Source: MarketDecode insider snapshot, June 7, 2026 (net buying value, 90d)
Marvell insiders
−$107M36 buys vs 80 sells, 90d
Lone buyer
TSM: 60 buys / 1 sellthe foundry, not the moat
What to watch
20-day line ~$208plus the Dec 1 print
Resolution window — 3 months
What would confirm or invalidate this read
Confirmation
By ~2026-09-07: (a) analysts RAISE Marvell price targets toward or above the ~$263 price (the highest published target is $200 today, all pre-June 2); (b) 30-day EPS revision breadth stays net positive (25 up / 3 down today); (c) the December 1 print / pre-announcements show the custom-silicon + interconnect ramp converting into accelerating revenue and raised guidance; (d) Marvell holds its 20-day line (~$208) and the connectivity complex (AVGO, CRDO, ANET, COHR) firms up.
Invalidation
By ~2026-09-07: (a) Marvell breaks and holds below its 20-day line (~$208) and gives back the Jensen pump; (b) 30-day revision breadth rolls over from 25 up / 3 down; (c) insider net selling accelerates beyond the −$107M 90-day pace; (d) the trillion-dollar narrative fades without analysts moving estimates — the price falls back toward the $130–$200 target band rather than the targets rising to meet it.