At 4:20 p.m. ET on Wednesday, August 26, one company published one quarter of numbers, and for a few minutes everything else on Wall Street was a footnote. Nvidia's fiscal second quarter: revenue of $96.2 billion, up 106% from a year ago. Data-center revenue of $89.0 billion — 92% of everything the company sells. Adjusted earnings of $2.22 a share, more than double last year's. And guidance of $108 billion for next quarter, roughly $4 billion above what analysts had penciled in. The market had spent the entire week positioned around this single release — trading desks called it, without much irony, the macro event of the week, in a week that also brings the Fed's preferred inflation gauge. This piece is not really about whether the numbers are good. (They are; at this scale they may be without precedent.) It's about what the night measured: how much of the modern market now runs through a single name — and what the history of markets that crowned a champion says about the years that follow.

+106%
Nvidia's year-over-year revenue growth in the Aug 26 report — $46.7B to $96.2B (company filing)
44pp
How far the cap-weighted S&P 500 has pulled ahead of the same 500 stocks equal-weighted since Jan 2023 (SPY vs RSP)
2.4×
How much likelier a big S&P 500 day (±1.5%+) is on the day after an Nvidia report than on any other day, 2023–2026 (21% vs 9%)
1 of 4
Alphixir strategy books holding Nvidia at the sealed 6:32 a.m. commit on the morning of the print — at 2.2%

The print, in one table

First, the facts of the report itself — because the scale is the story. A company already worth around five trillion dollars did not grow like a mature blue chip; it grew like a startup.

MetricQ2 FY2027 (qtr ended Jul 26, 2026)A year earlierChange
Revenue$96.2B$46.7B+106%
Data-center revenue$89.0B$41.1B+117%
Data center, share of total revenue92%88%
EPS (adjusted)$2.22$1.05+111%
Next-quarter revenue guidance$108B ± 2%vs. consensus ≈ $104B
Source: Nvidia's Q2 fiscal 2027 earnings release, August 26, 2026. Consensus figures as widely reported ahead of the print. EPS is the adjusted figure against an adjusted year-ago base.

Read the second and third rows again. Ninety-two cents of every revenue dollar now comes from data centers — which is to say, from one theme: AI compute build-out. The largest holding in the S&P 500 is, functionally, a pure-play on a single trade. If you own an index fund, you own that trade — sized for you by market capitalization, whether you chose it or not. That is not a criticism of the company. It is a description of the index.

Concentration, measured without an opinion

You don't need anyone's opinion to measure how concentrated this market is. Two ETFs hold the same 500 stocks: SPY weighs them by market value, RSP weighs all of them equally — Nvidia and the 500th company alike. Any gap between the two lines is, by construction, the part of the rally that came from the big getting bigger.

100150200Jan 2023 = 1002023202420252026SPY +110%RSP +67%
SPY · cap-weighted S&P 500RSP · the same 500 stocks, equal-weighted
The same 500 stocks, two ways of weighing them, indexed to 100 in January 2023 — the start of the AI rally. The gap between the lines is a direct measure of how much of the advance ran through the largest names. Source: Tiingo daily adjusted closes, SPY and RSP.

Since January 2023, the cap-weighted S&P 500 has returned about +110%; the equal-weighted version of the exact same stocks, about +67%. A 44-point gap, opened in less than four years. By 2025, the ten largest names accounted for roughly 40% of the index's value — about as top-heavy as the S&P 500 has been in over half a century of records. Honesty requires the other side too: this concentration is not, by itself, a bubble reading. Unlike the class of 2000, today's giants earn enormous and growing profits — $89 billion of quarterly data-center revenue is cash, not a promise. Concentration of price has tracked a genuine concentration of earnings. Both things can be true: the rally has been real, and it has been narrow.

One income statement, everyone's trading day

Here is the part that would have sounded absurd a decade ago: the day after Nvidia reports is now handicapped, hedged and traded as a whole-market event. We measured what that actually amounts to across the 14 Nvidia reports from February 2023 through May 2026 — and the honest answer has two halves. The typical reaction day is surprisingly ordinary: the index's median move on those days was 0.55%, barely above the 0.50% of a normal day, even while Nvidia itself swung a median 4.4% overnight — the market largely pre-positions, and the print often resolves into rotation rather than direction. But the tail is where the concentration shows: a big index day (±1.5% or more) has been 2.4× likelier the day after an Nvidia report than on any other day. On February 22, 2024, the index rose +2.1% — its best single day in over a year, and nothing about the other 499 companies had changed overnight.

-1%0%+1%+2%median S&P day ±0.50%2023−1.4%+2.1%2024−1.6%2025−1.5%2026
The S&P 500's close-to-close move (SPY) on the trading day after each of Nvidia's 14 earnings reports, February 2023 – May 2026, against the median move of all other trading days (dashed). These are index days, not Nvidia days: one company's income statement repeatedly setting the tape for 500 stocks. Source: Tiingo daily adjusted closes; report dates from Nvidia's public releases.

Note the bars point both ways — the index has been dragged down by these prints as well as up. And as we publish, the 15th bar does not exist yet: Wednesday's report landed after the close, and the reaction session hasn't opened. Whatever it brings, one line item — data-center revenue — will have more to say about what kind of day 500 stocks have than anything those 500 companies do themselves.

What happened after markets crowned a champion

None of this is new. Every era of this market has had a stock that was the market — the one name whose results everyone waited on. The mainframe era had IBM, the largest stock in the index for a generation. The PC era had Intel. The internet build-out had Cisco, which in March 2000 briefly became the most valuable company on earth — the picks-and-shovels supplier of the last great capex boom, which is exactly the role Nvidia plays in this one. Before all of them, the early-1970s "Nifty Fifty" taught the same lesson in plural form. Here is what the crown cost, computed from the price record:

The championWhen one name was "the market"Peak → troughUnderwater
IBM
the mainframe era
The largest stock in the index for a generation; computing itself, as far as the market was concerned−69%
Aug 1987 → Aug 1993
9 years
back to even Oct 1996
Cisco
the internet buildout
Briefly the most valuable company on earth in March 2000 — the picks-and-shovels of the last great capex boom−89%
Mar 2000 → Oct 2002
21 years
back to even Aug 2021
Intel
the PC era
The chipmaker whose product cycle set the tone for the entire tech tape of the 1990s−82%
Aug 2000 → Oct 2002
18 years
back to even Mar 2018
Nvidia
the AI era
92% of its revenue is one theme — AI data-center capex — and its report day moves the whole index (chart above)open question
Each era's dominant stock, its peak-to-trough drawdown after leadership peaked, and the time until a buy-and-hold investor was made whole, dividends reinvested. Computed from Tiingo adjusted closing prices; drawdown windows: IBM 1987–1993, Cisco and Intel 2000–2002. The Nvidia row is the question, not an answer.

The detail worth sitting with is not the drawdowns — it's that the businesses mostly kept delivering. Cisco's revenue was higher in 2004 than in 2000; Intel remained hugely profitable straight through its 82% drawdown; IBM's machines kept running the world's back offices. The companies were fine. The prices had assumed a future so large that even good execution couldn't fund it. That is the two-sided nature of concentration: the same weight that pulls the index up on the way up makes the index hostage on the way down — and the unwind, when it has come, has historically taken years, not quarters. We've charted the largest of those unwinds in our piece on the dot-com crash. To be equally clear about the differences: Nvidia's earnings are real and enormous in a way Cisco's 2000 valuation never was. History offers a base rate here, not a verdict.

What we committed before the number existed

Our own small corner of this story. Alphixir's four AI strategies seal their virtual books at 6:32 a.m. ET every trading day — published, hash-anchored, never edited after. Wednesday's seal ran about ten hours before Nvidia's release crossed the wire. What the books committed, on the record, before the number existed: the aggressive book swung back to 91.1% deployed and expressed the AI-chip trade through TSMC at 12.4% — the manufacturer, not the designer. The core book sized TSMC at 10.3%. And Nvidia itself — the one stock the entire market was waiting on — appeared in exactly one of the four books, at 2.2%.

That is not a call on the quarter; our engine doesn't forecast earnings, and we've written about why the day-after gap is close to unpredictable. The point is subtler. In a market this concentrated, any portfolio that sizes positions by its own rules — conviction, volatility, a half-Kelly cap — rather than by market capitalization is making an implicit bet against concentration every single day, whether it means to or not. Our books hold the market's largest company at 2.2% in one book and zero in three. If the champion keeps compounding, that stance will cost our virtual books against the index, publicly. If concentration mean-reverts the way it has before, the same stance will show up on the other side of the ledger. Either way, it was committed before the number existed and graded at the close after — and Thursday's 6:32 a.m. seal, the first one that gets to react to the print, will be on the public record too.

Takeaway

1. Concentration can be measured without an opinion: the same 500 stocks, cap-weighted versus equal-weighted, have split by 44 points since January 2023. The rally is real, and it is narrow — both at once.

2. One company's report day doesn't always move the market — the median reaction day is ordinary — but it makes a big index day 2.4× likelier, and it handed the index its best day in a year in February 2024. When the largest holding is a pure-play on one theme, its earnings print carries macro-sized tail risk, in both directions.

3. The base rate from past champions — IBM, Cisco, Intel — is drawdowns of 69–89% after leadership peaked, and 9 to 21 years underwater even with dividends reinvested, usually while the underlying business kept growing. That is a pattern about prices, not a prediction about Nvidia. Nobody reliably times it — which is why we publish commitments before outcomes, not forecasts.

Informational and entertainment content only. Not investment advice. Nvidia figures are from the company's August 26, 2026 earnings release; price data from Tiingo (SPY, RSP and individual stocks); sealed book figures from Alphixir's public daily record — virtual portfolios, no real money. Company names appear as factual and historical description, not as recommendations. Nothing here is a view on where any stock or the market is going.

Informational and entertainment content only. Not investment advice. Charts and tables are built from real price data (Tiingo) and Alphixir's own sealed daily record — virtual portfolios with no real money — and refresh when this page is regenerated.