Marvell tripled its way into the index. Now the index has to buy it.

On June 22, Marvell Technology joins the S&P 500, and on that day every fund built to track the index will buy it. Not because a portfolio manager studied the chip designer and judged the price fair. Because the rules require it. Marvell will replace Pool Corp, a swimming-pool-supply distributor, in the quarterly reshuffle that S&P Dow Jones Indices announced on June 5; Flex takes the seat the Campbell's soup company vacated. The symbolism — silicon displacing soup and pool floats — writes itself, and I would rather you ignore it. The interesting thing is not that chips are replacing chowder. It is that roughly ten trillion dollars now has to own a stock it was never asked to have an opinion about.

Let me say the obvious part first, because skipping it is how people misread columns like this one. Marvell is a real company. It designs the high-performance silicon — the custom accelerators, the networking and interconnect parts — that the AI data-center build-out actually runs on, and unlike a great many names riding the same wave, it makes money doing it. Index membership is not a participation trophy: one of the entry criteria is four consecutive quarters of positive earnings, and Marvell cleared that bar on the back of demand that is plainly real. Jensen Huang stood on a stage at Computex and called it "the next trillion-dollar company." When the man selling the shovels names your firm the next great mine, that is worth noting. So nothing here is an argument that Marvell is a bad company. The argument is narrower, and I think more useful: a good company and a good price are different claims, and the machine that takes over on June 22 is built to ignore the difference.

The index doesn't get to ask the price

Here is the mechanism, because the mechanism is the whole story. A fund that promises to track the S&P 500 is promising, in effect, to hold every name in the index at its index weight — whether the stock is cheap, dear, or absurd. When S&P adds a name, those funds do not get to deliberate. They have to buy, near the close, around the effective date, at whatever price the market is asking that day. The fund's job is not to be right about Marvell. Its job is to not deviate from the index. Price is an input it has contractually agreed not to care about.

Multiply that indifference by the money playing the game and you can see the size of it. Passive funds overtook active ones in US assets this year — call it roughly twenty trillion dollars that has decided, with some justification, that stock-picking is a mug's game. Of that, something close to ten trillion tracks the S&P 500 directly. The three largest S&P 500 ETFs alone hold nearly $2.7 trillion between them, and Vanguard's just crossed a trillion on its own. Behind them sit the target-date funds, the pensions, the 401(k) defaults, all benchmarked to the same list. When the list changes, all of that money moves the same direction at the same time for the same non-reason. The mechanical buying in a name Marvell's size runs into the billions. None of it is a vote of confidence. All of it looks exactly like one.

The fund's job is not to be right about Marvell. Its job is to not deviate from the index. Price is an input it has contractually agreed not to care about.

I have seen this film, and I keep the ticket stubs. The cleanest case is Tesla, added to the S&P 500 in December 2020 — at the time the largest inclusion the index had ever had to absorb. Between the announcement and the inclusion date the stock ran up something like seventy percent, because everyone who understood the mechanism front-ran it, buying ahead of the funds that would be forced to buy regardless. The index funds then dutifully paid the elevated price on behalf of holders who never chose Tesla and mostly never noticed. The stock was not wrong to be in the index; it was a real company making real cars. The price the index paid was a separate matter, and anyone who bought the inclusion pop spent the next year underwater. Being in the index and being a good buy on inclusion day are not the same event, and the gap between them is paid for by people who think indexing means they have opted out of making the bet.

Go back further, into the file I keep of every "this time is different," and the pattern gets less flattering. In 1999 and 2000 the committee added the names the era adored — Yahoo, JDS Uniphase, the optical and broadband darlings — at or near their peaks, because that is precisely when those companies grew large enough to qualify. Market-cap-weighted indexing is, by construction, a momentum strategy wearing the costume of neutrality. It adds things after they have gone up, in proportion to how far they have gone up, and it does so with other people's retirement money. The committee is not forecasting. It is ratifying. And the ratification arrives, reliably, after the move.

A bad company or a bad price

So apply the only question that has ever protected anyone in a market like this one: are we worried about a bad company or a bad price? With Marvell I am not worried about the company. The AI-infrastructure demand is real, the custom-silicon franchise is genuinely good, the earnings are genuinely there. If you had asked me three years ago whether Marvell would end up an S&P 500 constituent riding a generational chip cycle, "yes" would have been a reasonable bet. The trouble is the second half of the sentence: the stock has roughly tripled this year. A company can be excellent and a stock can be expensive at the very same moment, and the most costly mistakes in my notebook all came from people who proved the first half to themselves and assumed it had settled the second.

Then there is the trillion-dollar line, which I distrust on principle. Round numbers are not analysis; they are narrative furniture. "The next trillion-dollar company" is a wonderful thing to say at a trade show and a dangerous thing to underwrite a position with, because a trillion is a destination chosen for its count of zeros, not derived from anybody's cash flow. When a target is round, ask who profits from your believing it. Huang profits — every dollar of Marvell's ascent is a dollar that validates the demand for Nvidia's own parts and the build-out the two of them sell into. That does not make him wrong. It makes him interested, which is a different thing, and a column's whole job is to keep the two apart.

Here is what genuinely worries me, and it is not Marvell in particular. It is that the mechanism I just described has quietly become the marginal buyer of the entire AI trade. When ten trillion dollars buys names in proportion to their size, and adds them only after they rise, the biggest winners receive the most forced buying, which makes them bigger, which earns them still more forced buying. That is a feedback loop, and feedback loops are glorious on the way up and merciless on the way down — because the same rules that compel buying Marvell near a high will compel selling it, in proportion, if it ever slips out of favor or out of the index. Passive money is sold as the absence of a bet. It is in fact an enormous, automated, price-insensitive bet that whatever has already won will go on winning. That is a fine wager right up until it isn't, and the people making it have agreed, in advance, not to look.

• Marvell joins the S&P 500 effective June 22, replacing Pool Corp; Flex replaces the Campbell's Company, in the index's quarterly rebalance announced June 5.
• Roughly $10 trillion tracks the S&P 500 directly; the three largest S&P 500 ETFs alone hold nearly $2.7 trillion, with Vanguard's VOO past $1 trillion on its own.
• Marvell's shares are up by something like 210% in 2026, and rose again on the inclusion news — the kind of move that qualifies a stock for the index in the first place.
• Index funds must buy at the prevailing price around the effective date, regardless of valuation: inclusion is mechanical demand, not a judgment that the price is sane.

I will say the thing I always say, because not saying it would be a cheat. I have been early before. I flagged the AI-capex math as a problem well before the market agreed, and being early is a way of being wrong that feels like being right until your profit-and-loss disagrees. So I am not telling you Marvell is about to break, or that June 22 marks a top, or that you should do anything at all with a stock on the strength of a column. I am telling you what is actually happening: a real company, at a tripled price, is about to be bought in size by people forbidden from asking whether the price makes sense, cheered on by a man who profits from the cheering, on a schedule the whole market can see coming. Every one of those facts can be true and the stock can still go up. They have before.

The useful discipline is to hold two ideas at once, which is exactly the move a bull market punishes. Marvell is a good company. Marvell on June 22 is a forced buy at a momentum price, paid for by money that has promised not to care. Those are not in tension; together they are the whole picture, and conflating them — deciding that because the first is true the second must be safe — is the specific error that the index, by design, can no longer help you avoid. The index doesn't get to ask the price. You still do. Compared to when, is where I would start. Compared to when.