The week in signals: Jul 13–17

By Maya Koeva · July 17, 2026

A glossy chrome desk calendar page with a glowing signal waveform running across its face, marking a week of calls against outcomes.

The week-end scorecard, the same way we did it for June and with the same rules. The score does not try to time the market. It grades who is talking about a name and which way they lean, weighted by how credible those voices have been. The useful question on a Friday is narrower: when the score and the crowd genuinely disagreed this week, who was right so far?

First the honest backdrop: this was a rough week for the crowd's favorite trade. The memory and AI-infrastructure names our tracked accounts have been loudest about since June all fell hard: Nebius down 18%, SanDisk down 16%, Penguin Solutions down 15%, Micron down 9%. Keep that in mind for everything below; a lot of this week is that one trade deflating.

Every price move runs from the July 13 close to the July 16 close, the last completed session before we pulled the data on Friday morning.

What the signal caught

Aehr Test Systems is the week's cleanest case, and the shape is worth pausing on. Going into its earnings report the room was small but strikingly one-sided: a thin trickle of bullish posts from credible accounts, and essentially zero bearish calls all week. Then the report landed on Tuesday evening and a wave of 35 bullish mentions arrived before Wednesday's opening bell. The stock did the rest, closing Wednesday up 29% on the week before settling back to +24% by Thursday's close. Of the week's 48 mentions, 85% leaned bullish, 2% bearish, and a majority came from trusted accounts.

CompanyScoreMentionsTrustedWeek move
Aehr Test SystemsAEHR8748Bull 85% · Bear 2%58%+24%
ASMLASML8229Bull 83% · Bear 10%45%+3%
NVIDIANVDA8623Bull 74% · Bear 13%61%+2%
Mentions
77
PositiveNegativeChatterNoise
Aehr Test Systems mentions on Quantral: a quiet, credible bullish lead-in, then the earnings wave that landed before Wednesday's open.

ASML traced a smaller version of the same shape, 83% bullish with nearly half the chatter from trusted accounts, up 3% on the week. And NVIDIA, the most trusted room among the week's big names at 61%, ground out a 2% gain while the rest of the AI trade sold off around it.

What it stayed cold on

The SpaceX autopsy we published on Monday did not need a sequel for long. The score has sat at 19, one of the lowest on the loud end of the board, and the room stayed bearish: 57% of this week's 42 mentions leaned negative. The stock slid another 6% to $131.11, which now puts the biggest IPO in history 38% below its mid-June top of $211.39.

Oracle is the other cold read that kept paying. After Monday's drop the room went overwhelmingly one-sided, 78% of the week's calls bearish, with the heaviest wave landing Tuesday into Wednesday. The score sat at 30 and did not reward the dip. The stock bounced on Wednesday, and Thursday took the bounce away: down 6% on the week.

CompanyScoreMentionsTrustedWeek move
SpaceXSPCX1942Bull 26% · Bear 57%19%-6%
OracleORCL3041Bull 15% · Bear 78%10%-6%

One name deliberately not in that table: IBM. It fell 25% this week, almost all of it in one brutal Tuesday session, and its score today reads a cold 26, but we claim no credit for that. Before the drop, IBM barely registered on the board, a handful of mildly bullish, low-credibility posts. The bearish wave arrived on Wednesday, after the fall. A low score that shows up after the damage is the crowd reacting, not the signal predicting, and the difference matters if you want scores you can trust.

The honest part

Now the section this format exists for. Nebius was the single loudest name on our board this week, 143 mentions, 70% of them bullish, a third from trusted accounts, and a score of 79. It fell 18%, most of it in one Thursday drop. SanDisk, the cleanest catch in the June scorecard when it rose 19% against a falling sector, gave a big chunk of that back: down 16% this week, with the room cooling from June's 68% bullish to 48%. Penguin Solutions, which we took apart on July 10, scores 77 and still fell 15% with the rest of the trade.

CompanyScoreMentionsTrustedWeek move
NebiusNBIS79143Bull 70% · Bear 14%34%-18%
SanDiskSNDK7565Bull 48% · Bear 31%31%-16%
Price
$171.77-19% in June+0% off the low
Jul 6Jul 16
Nebius daily close: a 22% slide from the July 10 high while the room stayed 70% bullish.

That is not the score being wrong about the crowd. The credible voices on Nebius really were bullish, and here is the part worth sitting with: after an 18% week, they still are. As of Friday morning the room remains net bullish, and the trusted accounts have not flipped; the bears who did show up this week are mostly low-credibility ones. The score reads who is talking and which way; it never claimed to call where a crowded trade unwinds in any given week. June showed how this can resolve: AST SpaceMobile carried one of the highest scores on the board, fell 27% with its sector, and clawed back a good part of the drop in the month's final sessions. If the credible bulls on Nebius are right, this week was the dip in a longer story. If they are wrong, the turn will show up first in the same place this scorecard reads from: who is talking, and which way. One week is a snapshot, not a verdict, and next Friday we will grade this one again.

The takeaway

Strip out the selloff and the week comes down to the disagreements. Where the score and the loud crowd parted ways, the score's side had the better week: a quiet, credible, one-sided room that got paid on earnings (Aehr), and two cold reads that kept falling (SpaceX, Oracle). Where the score and the crowd agreed loudly, the market did not care, and the loudest bullish rooms took the hardest hits (Nebius, SanDisk). That is the read we think is worth having: not where the noise is loudest, but where the credible crowd is leaning, and whether the price is starting to agree.

For the method behind these calls, see how a track record is graded and volume versus signal.


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.