Nvidia’s quarter was not the problem. Revenue of $96.2 billion, up 106% year over year, earnings of $2.22 per share against $1.05, gross margin at 75.0% versus 72.7%, and a third-quarter guide of $108 billion plus or minus 2% against $57.01 billion a year ago. Colette Kress then told analysts to expect fiscal 2028 revenue growth of 70% while the street was carrying 44%, and added that customer forecasts point to demand roughly doubling, with the guidance reflecting supply constraints rather than orders. The stock rose about 6%. Marvell, Micron, SanDisk and Western Digital did not.

Marvell traded down roughly 6% Friday, two sessions after the Google disclosure closed it up 9.85% on Wednesday and another 1.6% Thursday. Nothing negative arrived in between. The 8-K is the same 8-K, the sell-side response has been uniformly constructive, and Reuters reported analysts reading the agreement as capable of adding more than 60% to the longer-term revenue outlook. What changed is who owns the shares.

The mechanics of the announcement explain most of the round trip. Marvell issued Google a warrant on 18 August covering 58,970,907 shares at an exercise price of $206.58, about 7% of shares outstanding and roughly $12.2 billion at the strike. Only about 1.36 million of those shares vest on a clock, in equal quarterly installments through the first year. The remaining 57.6 million vest as Google spends: 240 tranches of roughly 240,042 shares each, one for every $500 million of custom-products revenue Marvell recognizes from Google and its affiliates, measured from 1 August 2026 through 29 January 2033. Two hundred forty times $500 million is $120 billion. That is where the headline number comes from, and it is a ceiling on vesting rather than a revenue forecast.

Marvell disclosed a commercial agreement with Google on Wednesday and the stock opened 13% higher at $243.66, extending a year-to-date advance that already ran above 150%. The headline number circulating is $12.2 billion — the value of a warrant Marvell issued to Google covering 58,970,907 shares at an exercise price of $206.58. That figure is the least informative thing in the filing. The number that matters is $120 billion, and it is not stated anywhere in the announcement. It has to be derived from the vesting schedule, and once derived it tells you exactly how large Marvell believes this relationship becomes.

Morgan Stanley’s upgraded CXL forecast is circulating as a Marvell thesis, and the exhibit it comes from does not support the weight being put on it. The chart stacks two lines, memory expansion controllers and switch silicon, and carries the combined market from roughly $50 million in 2025 to about $4.05 billion in 2030. Both halves were raised hard. MXC went from $990 million to $2.1 billion. Switching went from $664 million to $1.9 billion. What did not change is the scale of the thing being doubled. Four billion dollars in 2030 is the entire industry, before Marvell divides it with Astera Labs, Montage, Microchip and Rambus.

The Philadelphia Semiconductor Index closed Friday more than 20 percent below its late-June record, the technical threshold for a bear market. Two trading sessions later, the index is up a combined figure north of 7 percent, with Tuesday’s 3.7 percent gain marking the second consecutive advance. The question facing the sector is whether this is the start of a genuine recovery or a bounce inside a larger correction.

What Broke the Rally

The selloff had three distinct triggers layered on top of each other. First, valuation exhaustion: chip names had posted triple-digit year-to-date returns before the reversal, leaving the trade crowded and vulnerable to any negative catalyst. Second, a competitive scare — a Chinese open-source model release narrowed the perceived gap with frontier US labs, raising questions about the compute intensity assumptions baked into AI infrastructure spending forecasts. Third, TSMC’s capital expenditure guidance came in well above prior estimates, which the market read as a margin-pressure signal rather than a demand-confirmation signal. Once the reversal began, systematic de-risking and algorithmic flow amplified the decline across both US and Asian listings, with SK Hynix and Samsung posting double-digit single-session drops in Seoul.

United Microelectronics Corporation, Taiwan’s second-largest contract chipmaker, announced the first mass-production wafer delivery of photonic integrated circuits from its Singapore 12-inch fab on July 14. The wafers were built in partnership with SILITH Technology, a Singapore-headquartered fabless silicon photonics company whose 1.6T platform has already shipped more than 8 million 100G and 200G-lane PICs. The two firms took the platform from development to production readiness in 18 months, and a leading cloud infrastructure customer has already qualified it for volume deployment.

On June 2, Jensen Huang turned to Matt Murphy on a Computex stage in Taipei and called Marvell the next trillion-dollar company. The stock rose 32.52% that day, its largest single-session gain on record, adding roughly fifty billion dollars in market value before the close. Every desk on the Street ran the clip. Almost none of them ran the number underneath it.

The number was disclosed six days earlier, on the May 27 fiscal first-quarter call, and it carried no theater at all. Marvell raised its forward guide by roughly five billion dollars and lifted interconnect growth from 50% to over 70% year over year. The trillion-dollar line moved the tape. The guide raise moved the thesis. Those are not the same event, and conflating them is how investors end up paying for sentiment while telling themselves they bought fundamentals.

When Jensen Huang stood onstage with Matt Murphy at Computex in Taipei and called Marvell the next trillion-dollar company, the market did not treat it as a courtesy. The stock posted its largest single-day gain on record, jumping more than thirty percent the following session. Endorsements from rivals are usually cheap. This one was not, because Huang was not flattering a partner. He was describing the part of the AI buildout he understands better than anyone, and naming the company that owns it.

CXL was sold as a capacity story: extend the memory pool past the DIMM slots soldered to the motherboard. Marvell’s argument with Structera is sharper than that. The pool itself is half-empty. The data sitting in DRAM is compressible, almost no CXL controller touches it, and Structera does — in dedicated silicon, at line rate, invisible to the host.

The number circulating is 3.64x, the top of the range Marvell cites for mixed real-world data types, which it claims match or closely approach what host-side LZ4 achieves in software. Field reporting has been more conservative; ServeTheHome quoted Marvell putting practical ratios at 1.8x to 2x. Both numbers point the same way. Even a flat 2:1 halves the effective cost per gigabyte of a memory pool, and memory is the single largest line item in that pool.

There is one argument running underneath every chip-stock target reset this week, and it is not really about chips. It is about whether memory, storage, and the wires between accelerators are commodity components that move on the old PC-and-mobile cycle, or mission-critical AI infrastructure whose demand scales with every model upgrade, every reasoning capability, and every agentic deployment.

Bank of America just answered that question with its wallet. On June 23 — a day the group was getting hit, not bid — Vivek Arya raised Micron to $1,500 from $950 and reframed DRAM and high-bandwidth memory as structural AI infrastructure rather than a cyclical good. The same desk lifted Marvell to $365 the same session and circulated a note arguing the broader memory-plus-interconnect complex represents another trillion-dollar opportunity for chip names. That is the tell. When one analyst makes the identical structural call across DRAM, NAND, and custom silicon on a down day, it is not a price target. It is a thesis.

Noam Mizrahi has been saying the same thing for two years, and the market has only just decided to believe him. Marvell’s corporate CTO, based at the company’s Israeli site and a Technion graduate, has argued since the early innings of the AI build-out that the constraint on the next leap was never going to be the processor. It was going to be the wire between the processors — and then the optics, when copper ran out of reach. The industry called this a backwater. Marvell bet the company on it. The bet has now compounded into one of the most violent re-ratings the semiconductor tape has produced this cycle.

KeyBanc’s John Vinh raised his Marvell price target to $385 from $260 on Thursday, a 48% increase, and kept his Overweight rating. The headline number is large. The argument behind it is more interesting, because it inverts the hierarchy that has carried the stock for two years.

For most of the AI cycle, Marvell has been priced as a custom-silicon story. The thesis was the XPU pipeline: bespoke accelerators for AWS and Microsoft, a clear line of sight to roughly $10 billion in custom-chip revenue by fiscal 2029, and a seat at the table next to Broadcom in the ASIC duopoly. Networking was the supporting act. Vinh has now promoted it to lead. He came out of recent investor meetings calling networking the most durable growth opportunity Marvell has, and put a number on the scale-up market — optical links, silicon photonics, and high-speed switching — of around $30 billion by 2030.

When Jensen Huang stood on the Computex stage and called Marvell a potential trillion-dollar company, it sounded like a courtesy extended to a new partner. It is not. It is a forecast with a visible arithmetic spine. Marvell closed near $325 in mid-June carrying a market capitalization around $272 billion. A trillion dollars is roughly 3.7 times that. The question is not whether the path exists. It exists, it is mapped, and Marvell’s own guidance lays most of the mileposts. The question is how many years it takes and what has to hold along the way.

The headline number is clean and the headline framing is wrong. Nvidia did not buy $2 billion of Marvell stock in the market. On March 31, 2026, it purchased two million shares of newly issued Series A Convertible Preferred Stock at a stated value of $1,000 each, a private placement that put $2 billion of fresh cash directly onto Marvell’s balance sheet. That distinction is the entire story. Nvidia did not become a passive holder of MRVL. It became a senior, structured creditor-equity hybrid with a conversion option struck deep below where the stock now trades, and it did so as the price of admission to a partnership designed to neutralize the single largest threat to its own franchise.

Thesis

The headline was a record: $2.418 billion in revenue, up 28% year-over-year, with $0.80 of non-GAAP earnings. The headline is not the story. The story is what management did to the out-year model. On the print it raised the fiscal 2028 revenue outlook toward $15 billion and the fiscal 2027 outlook to approach $11 billion, and it did so on bookings rather than hope, citing AI-related order momentum it called exceptional. Marvell is no longer a diversified chip vendor with an AI option bolted on. It is a custom-silicon and interconnect company whose addressable market is being rewritten by the hyperscaler decision to design proprietary accelerators and buy the connective tissue around them.