Marvell (MRVL) to $1 Trillion: Nvidia's Road Requires 70% Gross Margins, Broadcom's Requires Stock Marvell Has Already Spent

Two semiconductor companies have gone from Marvell’s current size to $1 trillion, and they did it by completely different mechanisms. Sorting out which mechanism is available to Marvell is more useful than compounding a revenue line forward, because the answer to how long it takes depends on which road the company is actually on. Marvell closed the earnings session at $241.42 with a market capitalization of $211.7 billion, up roughly 184% year to date against a 52-week range of $61.44 to $329.88, and traded to $224.34 after hours on a print that beat and raised. It needs 4.72 times from here.

Nvidia’s road was margin leverage. It went from under $300 billion to $1 trillion in roughly eight months during 2023, and the mechanism was not revenue growth in the conventional sense. It was that incremental revenue arrived at gross margins above 70% against a fixed cost base that had already been built, so operating income grew several times faster than the top line and the market re-rated the earnings power rather than the sales. That is the fastest path to a trillion dollars ever run in this industry, and it is structurally unavailable to Marvell. The third quarter guide implies incremental gross margin near 52% at the midpoint of the 57.5% to 58.5% band, below the 58.9% just delivered, and management named the reason: custom silicon acceleration is a sequential margin headwind. Marvell’s growth engine dilutes its own margin. Every dollar of the custom revenue that gets the company to scale arrives worth less than the dollar before it. You cannot produce Nvidia’s re-rating out of a mix that moves the wrong way, because the entire re-rating was a bet on operating leverage that a rising-margin mix made obvious and a falling-margin mix makes speculative.

Broadcom’s road was consolidation, and this is the argument against my own earlier arithmetic. Modeling Marvell’s path as $18 billion of fiscal 2028 revenue compounding organically at 35% or 40% assumes the Broadcom playbook is closed. It is not. Avago was a roughly $4 billion company at its 2009 listing. It reached $1 trillion by buying LSI for $6.6 billion, Broadcom Corporation for $37 billion, Brocade for $5.9 billion, CA Technologies for $18.9 billion, Symantec’s enterprise business for $10.7 billion and VMware for about $61 billion, in a sequence that added well over $140 billion of purchased assets across fifteen years. Organic compounding was the smaller contributor. If the question is how long it takes Marvell to reach $1 trillion, an acquisitive path can compress that timeline by years, and Marvell has already demonstrated it will use the tool. Inphi cost $10 billion in 2021 and became the optical DSP franchise that is now the most defensible thing the company owns. Cavium, Innovium, Aquantia and Avera preceded it. Celestial AI closed this year at $3.25 billion, $5.5 billion with the earnout, and XConn closed in February.

Here is the complication that makes the Broadcom road expensive rather than closed. Hock Tan’s acquisitions worked because Avago’s stock was accretive currency: it bought earnings at a lower multiple than the multiple its own shares carried, so every deal shrank the blended multiple and the arithmetic compounded. Marvell trades at 53 times forward earnings and roughly 37 times the fiscal 2028 number management just guided. On that measure its stock is the most valuable acquisition currency in the sector, and the logic says go shopping immediately. Except that the currency is partly committed. The Google agreement issues 58,970,907 shares at $206.58, close to 7% of the company, in 240 tranches of about 240,042 shares against each $500 million of custom product revenue through fiscal 2033. Marvell has already assigned a seven percent block of its equity to customer acquisition. Stacking acquisition currency on top of that means the share count grows on two fronts against the same denominator.

Which surfaces the thing that makes market capitalization milestones a poor way to think about this. A $1 trillion market capitalization is dilution-agnostic. Issuing shares makes it easier to reach, not harder, because it lowers the price per share required. Full vesting of the warrant takes the required price from $1,140 to $1,069. If Marvell then bought $150 billion of assets with stock, it would clear a trillion dollars of market capitalization considerably sooner and existing holders would have made nothing from the difference. The milestone and the return are separate questions and only the second one matters. Broadcom passes that test because the purchased earnings were bought cheaply enough that Avago holders were made better off despite enormous dilution. The test for Marvell is identical and has nothing to do with the trillion: whether stock issued at 37 times forward buys earnings at less than 37 times forward. Nothing about reaching a round number tells you the answer.

The moat argument sits underneath both roads and does not move. SerDes, optical DSP, the 800G and 1.6T interconnect ramp guided above 70% growth, the 51.2T switching franchise, and scale-out switch revenue projected past $600 million constitute the attach business that survives a hyperscaler bringing its compute die in house. That is what deserves the premium and it is what Inphi actually bought. The compute socket is the business large enough to matter at trillion-dollar scale and the business where Marvell has no structural defense against either Broadcom or a customer’s internal team. An acquisitive path does not resolve that tension, it just changes who is buying whom. It is worth noting that the assets Marvell would need to buy to reach $145 billion of revenue are, at current sector multiples, more expensive relative to their earnings than anything Hock Tan bought between 2014 and 2019.

Base case remains fiscal 2033 to fiscal 2034 with the stock in a $220 to $270 range into the October 6 investor day, on 35% to 40% organic growth into a high-twenties terminal multiple. The bull case moves to fiscal 2031 and requires either a sustained 50% growth rate with the multiple defended above 35 times, or a large accretive acquisition paid for in cash and debt rather than shares, with October 6 as the near-term catalyst if it splits the socket count into compute versus attach. The bear case is cohort derating: if the custom silicon complex converges toward the hardware multiples the sector has historically paid, Marvell moves toward Broadcom’s 20 times, the revenue required for a trillion moves to roughly $145 billion, and the target recedes faster than the company grows into it. Against 38 covering analysts, no sells, an average target near $257 and a high of $400, the most bullish published number on the street implies $351 billion, which is 35% of the distance.

The decision-relevant disclosure is the consideration mix on the next acquisition. Cash and debt says management believes the organic path works and is unwilling to dilute into it. A stock-funded deal at this multiple says management is running the Broadcom playbook, which is the faster road to the headline and the one where the milestone and the shareholder return come apart.