On 4 June, S&P Dow Jones Indices told the market it would not bend. After a formal consultation on whether to fast-track "MegaCap" newcomers into its flagship indices, the committee left the S&P 500's eligibility rules untouched: the twelve-month seasoning period, the minimum float, and the financial viability screen that has guarded the index for decades. That screen is simple. To be eligible, a company needs positive GAAP net income in its most recent quarter and across the most recent four quarters summed. No exceptions for size. Four days later, OpenAI confirmed it had confidentially filed to go public, with a target valuation above US$1 trillion. The largest US listing in years is heading straight for the one major index that will not take it.

The pitch

The case for changing the rules, pressed by bankers and lawyers through late 2025 and into this year, runs like this. A company worth a trillion dollars is, by definition, a large slice of the US equity market. An index that claims to represent that market and leaves the slice out is failing its own mandate. So the rules should flex to let the giants in quickly. And the rest of the index world agreed. Nasdaq rewrote its methodology, effective 1 May, to allow fast-track entry for big IPOs, with a float multiplier that lets a megacap qualify even at a thin free float. FTSE Russell ran its own consultation on dropping float minimums for the same reason. The momentum pointed one way, and S&P was widely expected to follow. It ran the consultation. It said no.

What it left in place is worth stating precisely, because the detail is the story. S&P did loosen the rules for its broad-market indices, the S&P Total Market and Completion indices, effective 8 June, so that a huge company with a tiny public float can now enter those. But for the S&P 500 itself, the financial viability screen stands. SpaceX, which is reportedly profitable, can get in once it clears seasoning, perhaps by mid-2027. OpenAI and Anthropic, which are not profitable, cannot, no matter how large they list.

What the rule is actually guarding

Strip away the index plumbing and the question underneath is who provides the exit. OpenAI is not going public because it needs the public's money. It raised US$122 billion privately in March at an US$852 billion valuation, the largest private fundraise on record, and it is burning something like US$27 billion this year. A listing is how early investors and employees turn paper into cash. Index inclusion is the quiet engine of that conversion. When a stock joins the S&P 500, the roughly US$13 trillion of passive money that tracks the index has to buy it, mechanically, regardless of what any fund manager makes of the company. VOO alone now holds about a trillion dollars; IVV and SPY add nearly another US$1.6 trillion between them. The profitability screen is the single tripwire that withholds all of that automatic demand until the company has earned a dollar.

That is the bet S&P is making, and it is not a bet about performance. It is a claim about what the words mean. "Representative of the US market" should still, in the committee's reading, carry an implication of being a viable business, not just a large one. The index will provide forced-buying exit liquidity to a company that has never made money only after that company stops being one.

The case that none of this matters

The strongest argument against reading too much into the "no" comes from people who model index flows for a living, and it deserves a fair hearing. The analyst who writes as bau laid it out before the decision landed: in performance terms, the profitability screen is close to noise. Measured against MSCI USA, which carries no such screen, the S&P 500 has drifted by about 13 basis points over fifteen years, with the two indices correlated at an r-squared of 0.999. Whatever the screen is doing, it is not meaningfully changing returns.

And the exclusion leaks badly. Unprofitable or not, OpenAI will still land in the Russell 1000, the Nasdaq 100, MSCI's and FTSE's global benchmarks, and S&P's own Total Market and Completion indices. By bau's estimate, S&P 500 membership is worth only about 12 to 13 per cent of the total passive demand a megacap IPO pulls in across all index families. Fortune made the political version of the same point, arguing that loosening the rules would have benefited SpaceX and Anthropic, not ordinary investors, which means that by holding firm S&P mostly spared index-fund holders from being conscripted into an unprofitable trillion-dollar stock. The protection is genuine, but it is small, and most of the money flows anyway. On this reading the whole thing is symbolism, and symbolism is cheap.

Why the symbol has teeth

This is where I part company with the it-does-not-matter view, on two grounds.

The first is that for this particular index, the symbol is the product. The S&P 500 is the benchmark whose name is shorthand for "the stock market" in retirement statements and evening news alike. An index whose authority rests on being the trusted mirror of corporate America cannot treat the contents of that mirror as a rounding error. The 13-basis-point argument proves the screen barely moves returns; it does not prove the screen barely matters, because the screen's job was never to boost returns. It was to define membership. Confusing the two is the central error in the dismissal.

The second is that the discipline is not costless theatre, and we know that because it has already been tested and has already cost the index money. Tesla was kept out of the S&P 500 until it strung together four consecutive profitable quarters, joining only in December 2020, by which point it was already one of the most valuable companies in the country and most of its historic run was behind it. The index swallowed that. It admitted one of the great wealth-creation stories of the era late, on principle, and wore the underperformance that came from missing the climb. That episode is the evidence that the rule binds rather than bends. When S&P now applies the same screen to OpenAI, you know it is not improvising, because you watched the same screen cost it dearly with Tesla and hold anyway.

There is a sharper objection buried in bau's own write-up, and it cuts toward S&P, not away from it. The waiver S&P rejected would have exempted only the very largest companies from the profitability screen. A US$1 trillion firm listing at 5 per cent float could have entered while a dozen profitable-margin US$100 billion companies stayed out on a technicality. The proposed rule, in other words, would have applied the principle to everyone except the names big enough to move the index most. Rejecting it was the consistent choice. If profitability is a real requirement, it cannot be the one rule that switches off precisely when the company is too big to question.

The bet

So here is the falsifiable version of all this. If OpenAI lists in September and grinds its way to four straight profitable quarters within a couple of years, the S&P 500 takes it, the June decision becomes a speed bump nobody remembers, and the dismissers were right that this was mostly about timing. If the profitability keeps receding instead, with cash burn near US$27 billion this year and no GAAP profit in sight, something more telling happens. The S&P 500 becomes the one benchmark that the investing public treats as its mirror and that never held the most-discussed company of the cycle. The distance between the market people picture when they say "stocks" and the actual AI buildout stops being rhetorical and becomes a line you can point to.

That line has a date attached, whenever it arrives: the first quarter OpenAI prints a GAAP profit, if it ever does. Everything between now and then is float arithmetic. Watch that quarter. It will tell you whether S&P drew a line the market eventually erased, or whether it drew the only line the AI buildout could not lobby its way across.