Revenue grew 36%. The stock fell 27%.

One game brought the crowd. It's thinning.

Grow a Garden, a farming game whose seeds grow while you're logged off, pulled Roblox to a record 151.5 million daily players last fall. Players have fallen three straight quarters since; hours played have fallen further still, down 27% from their peak.

Another take, while we're counting heads:

Flags 2 and 3 were Roblox's own decisions, and the filings talk about both — plus the bet behind them. The operative words, highlighted:

That bet came due on July 30, when Roblox had to put a number on next quarter.

The report was fine. The forecast wasn't.

Guidance is a company's own prediction. For seven straight quarters, the money that came in beat Roblox's own forecast. This time Roblox expects 14–18% less new money than the same quarter last year — the Grow a Garden quarter. It is the first forecast drop the company has ever given. Share prices are bets on what happens next: the forecast did the damage, not the quarter.

Roblox normally forecasts the next quarter and the year. On July 30 it gave only the quarter. Past September, Roblox itself won't say.

The next morning the stock fell 27% — the number the page opened with.

The losses are on paper. The cash is in the bank.

The crash pulled an old take back out:

That one, a chart can settle:

Six quarters, six red slips, $1.5B of “losses” — while $2.2B of leftover cash poured in and the level never once dropped. Both are true, because the reports count old sales slowly and count shares paid to staff as a cost: a minus on the slip, no money out of the jar. In the bank: $6.1B.

The money coming in got cut. The money going out didn't.

Losing money on paper never hurt Roblox; the cash pile kept growing anyway. AI is the first cost big enough to stop it: by Roblox's own forecast, next quarter's spare cash is roughly zero.

Roblox never reports a line called AI spending. No company has to. It reports where the money lands: buying the machines (cash leaving) and running them (a quarterly expense). Roblox says AI pushes both.

Buying is lumpy: almost nothing for two quarters, then a data center lands. Running only goes up. The one number Roblox gave: $170M of equipment in the September quarter, roughly eight times June's. That is why cash flow is forecast between −$60M and +$5M.

Roblox has spoken about 2026 three times. February: a plan for the money coming in, and a $470–520M machine budget. April: the money plan cut by about $1B, while the machine budget, in the same letter, went untouched. July: no year numbers. The scoreboard's third column is that same July 30 silence, seen from the spending side.

The crowd fell 19%. The stock fell 74%.

Step back, and the whole story fits one chart. The stock topped out at $141.56 on September 29, 2025; the players in the quarter ending the next day.

Since that peak the crowd is down 19% and the stock 74%, about four times as far. Nothing says one caused the other. But a share price is a bet on what the headcount does next; the gap is that bet being marked down.