# Retail never stopped buying SpaceX. Today the float doubles. > Retail has net-bought SpaceX every single day since the IPO, and today roughly 912 million locked-up shares come free. Quantral read the name cold for a month at a score of 19. After the first earnings report it flipped to 77, while the raw crowd stayed two to one bearish. The gap between those two readings is the whole story. By Maya Koeva · 2026-08-07  Since SpaceX went public on June 12, individual investors have net-bought the stock on every single trading day, without one session of net selling, while the share price lost more than half its value from the peak. Vanda Research, whose numbers [Yahoo Finance charted this week](https://finance.yahoo.com/markets/stocks/article/spacex-stock-got-cut-in-half-but-retail-investors-keep-piling-in-chart-of-the-day-100000031.html), puts $405 million into the first five sessions alone. On Wednesday, in the opening hour of a session that would end down 13.6%, retail bought another $22.7 million. That was more than three times the normal opening-hour pace and the third-biggest opening hour in the stock's 37-session life. Today the other side of the trade arrives. Roughly 912 million pre-IPO shares clear their lockup, which could double the public float overnight. Every share that has wanted out since June and could not get out is now free to leave. We wrote about this name a month ago, when Quantral's score [read it cold at 19](/blog/signal-autopsy-spacex) while the price was still $139. The stock has since fallen to $114.92. The score has gone the opposite way. ## What changed on Tuesday night SpaceX filed its first earnings report as a public company after the close on August 4, and the top line was not the problem. Revenue came in at [$7.8 billion, up 92%](https://fortune.com/2026/08/04/spacex-revenue-surges-92-to-7-8-billion-blowing-past-wall-street-expectations-by-nearly-1-billion/), beating consensus by nearly a billion dollars. Starlink doubled its subscriber base to 12 million at a steady $66 a month. The net loss shrank to $541 million from about a billion a year earlier. Then there was the other number. Capital expenditure hit $18.4 billion for the quarter, $15.8 billion of it on AI, an annualized run rate of $73.5 billion against consensus nearer $48.7 billion. The stock fell more than 7% after hours, then 13.6% the next session, then bounced 6.1% on Thursday to close at $114.92. Traders read the capex line, not the revenue line. ## The score went the other way Quantral's 7-day signal score on SPCX now reads 77, with the 24-hour window at 70. A month ago it was 19. Most of that move happened in three days. The raw mention count did not move that way at all. Over the seven days to this morning, the accounts we track produced 209 subject mentions of SPCX from 153 different accounts, and the directional ones ran 98 bearish to 55 bullish. Counted as a show of hands, the room is still nearly two to one against the stock. The score disagrees with the show of hands because it is not counting hands. Two things separate it: **The platforms split cleanly.** Of the bearish mentions, 85 came from Reddit and 13 from X. Of the bullish ones, 32 came from X and 23 from Reddit. Usually the two crowds rhyme. On this name, in this window, they are having opposite conversations. **The two sides are posting different things.** Among the bullish mentions, 11 are theses: dated, falsifiable arguments with a horizon attached, ten of them on X. Among the 98 bearish mentions, five are theses, and 83 are not calls at all, mostly one-line reactions in comment threads. The model weights a stated argument from an account with a graded record far more heavily than a throwaway, and the top of the weighting is dominated by long-horizon bullish posts from accounts with hundreds of graded calls behind them. The heaviest single contributor this week carries about a third of the directional weight on its own. The number is doing what we built it to do. Volume sits on one side of this stock and dated arguments from graded accounts sit on the other, and when those two come apart, the score follows the arguments. *[Chart: SpaceX (SPCX) mentions on Quantral, July 13 to August 6. The bearish column runs heavier through most of the stretch. August 4, earnings day, is the busiest day SPCX has had since it listed, at 128 mentions split 36 bullish to 60 bearish. The price drop landed the following session.]* ## The tape has not agreed *[Chart: SpaceX daily close, July 13 to August 6: a steady 22% grind down into the July 31 low, a run into the first earnings report, then the post-earnings whipsaw of minus 13.6% and plus 6.1%.]* From the June 16 close of $211.39, SPCX is down 45.6%. It sits 14.9% below its $135 first-day close, and 17.4% below where it was when we last wrote about it. The cold read from July tracked a real decline. The warm read is three days old and has proved nothing. We would rather say that plainly than dress it up. The score reads the conversation, and it does not forecast the price. Today it says the argued case for SpaceX sits on the bullish side while the reflexive case stayed bearish. Whether the argued case holds up is a separate question, and today the market starts answering it. ## Why today is the test The lockup is the cleanest event this stock will get for a while. Vanda's read on retail is that they "continue to see SPCX as a transformational AI story, as opposed to a space exploration or interplanetary travel stock," which is the thesis in the bullish posts driving the score: $2.6 billion of AI segment revenue growing 247%, $14.1 billion in new cloud contracts, 1.4 gigawatts of compute. The bearish posts have pointed at the float and the capex since June, and in July that argument was right. Both sides now have a date. If buyers absorb the unlock, the accounts calling the float fear overdone earn a mark in their column and the score's flip looks early. If 912 million shares find no bid, the one-liners in the Reddit threads were the better read, and the score will follow the credible accounts back down as they concede. Either way the number moves for a reason you can go and read. ## The takeaway A month ago this name showed the score refusing to follow fame down. Today it shows the less comfortable half of the same design: the score changes its mind on evidence, before the price agrees, and it can be early or it can be wrong. It weighs who is arguing what, and it moved because the arguing changed. Keep the mechanism rather than the call. A raw mention counter looking at SPCX this week would report a stock the crowd hates, which is true and thin. [Credibility weighting](/learn/what-is-a-credibility-score) reports something narrower: the accounts posting dated theses went one way, the accounts posting reactions stayed the other way, and the gap between them is now the size of the whole read. You decide what to do with that gap. Today the market starts scoring it. --- *Quantral surfaces signals and context from public sources to support your own research. Nothing here is financial advice or a recommendation to buy or sell. Past signals are not indicative of future results.*