# Nvidia cleared the bar. Grading the room that called it. > Yesterday we published the room's read on Nvidia hours before the print: crowd 3-to-1 bullish, the tracked accounts at 6-to-1. Nvidia beat on everything and guided nearly $4 billion above consensus. The grade, the caveat from May, and the one chart that shows who showed up after the fact. By Maya Koeva · 2026-08-27  Yesterday morning, hours before Nvidia's second-quarter print, [we put the room's read on the record](/blog/nvidia-earnings-sentiment): 386 mentions across the accounts Quantral tracks, the crowd a shade over 3-to-1 bullish, and the tracked accounts, the ones with real graded records, at 62 to 11, nearly 6-to-1. We committed to grading that read after the print either way. This is the grade. ## What the print delivered Nvidia cleared every bar it was given. Adjusted earnings came in at [$2.22 per share on $96.2 billion in revenue](https://fortune.com/2026/08/26/nvidia-results-q2-earnings/), against the roughly $2.09 and $92 billion the street expected, with revenue up 106% from a year ago. Data-center revenue hit $89 billion. Guidance did the heavier lifting: about [$108 billion for the third quarter](https://www.tikr.com/blog/nvidia-reports-q2-earnings-august-26-the-q3-guide-and-china-decide-the-stock), nearly $4 billion above the $104.2 billion consensus, and that number assumes no data-center compute revenue from China. The same evening, reports landed that Nvidia [agreed to buy Hugging Face for $12.9 billion](https://techstartups.com/2026/08/26/nvidia-agrees-to-buy-hugging-face-for-12-9-billion-in-major-ai-deal-taking-control-of-the-github-of-ai/), the distribution hub for open-source AI models. The tape's first verdict was messier than the numbers. Nvidia had closed Wednesday's regular session at $209.66, down another 1.6%, its eighth red close in nine sessions. The stock [dipped on the release, then reversed to gain about 4.7% in after-hours trade](https://www.benzinga.com/markets/tech/26/08/61452353/nvidia-earnings-blowout-after-hours-stock-dow-futures-jensen-huang) as the call went on. Call it just under $220 as of this morning: about half the seven-red-day slide recovered, still below the $225.30 where the slide started two weeks ago. ## Question one: the room's lean, graded The going-in piece framed the print as a test between two groups. The whole crowd ran 3.2-to-1 bullish but day-traded its own opinion, flipping net bearish on Monday's seventh red close and flipping back on Tuesday's bounce. The 21 tracked accounts with a [credibility score](/learn/what-is-a-credibility-score) above 0.5 and at least ten graded calls went 62 to 11 and did not move all month. On the quarter itself, the tracked room was right. The specific bull case running through those mentions, demand outrunning supply into a strong quarter, compressed valuation, buyers treating the red streak as an entry, is the case the print confirmed. The most repeatable bearish argument, that Nvidia tends to move less than option buyers hope and sells off even on beats, got the first half right at 4:20pm and lost it by the end of the call. So the accounts that decided weeks ago and let the tape disagree got the quarter they described. The accounts that repriced their opinion twice in two days off single red and green closes got it too, but only by accident of the final flip. ## The chart that shows who showed up late The same mentions chart from yesterday's piece, extended by one bucket. The last bar is the overnight reaction, counted through early Thursday morning. *[Chart: Nvidia mentions on Quantral, August 6 to 27: three weeks of debate, then the answer arrives and the biggest bar of the month arrives with it. The final bar is the post-print reaction, counted through early Thursday morning ET.]* The overnight bucket holds 224 subject mentions, 146 bullish to 31 bearish. That single after-the-fact bar is more than three times the biggest bar of the entire three-week debate. Bullishness got loudest the moment it stopped requiring nerve. Most of the overnight surge came from X, 177 mentions across 45 handles, with 47 more from the subreddits we track. The tracked accounts joined the reaction too, at 42 bullish to 6 bearish overnight. The difference: they were saying the same thing last week, into the seventh red close, when saying it cost something. CKCapitalxx, who went in 3-to-2 bullish and mixed, came out of the call with seven straight bullish mentions and a buy alert on fundamentals catching up. [jukan05](/blog/who-is-jukan05) moved from the print to the two new storylines, the Hugging Face deal and Nvidia's move to lock up memory supply. ParadisLabs called it the cheapest high-quality growth story on the board. The six overnight bearish mentions point at the one soft spot in the print, margins: Nvidia [guided gross margin down toward 71 to 72% by early next year](https://financefeeds.com/nvidia-nvda-q2-fy27-earnings-results/) as memory prices climb, and jasonschips and ParadisLabs both flagged what that costs. PegasusFund, one of the two named bears going in, still questions how the long-term estimates get hit. That is the whole argument for reading the room with a credibility layer, compressed into one bar. Anyone can be bullish on Thursday morning. The reading with value was Tuesday's, and it was on the record with names and track records attached. ## Question two: the score Through seven straight red closes, Nvidia's score on Quantral refused to follow the price down, sitting at 75 the morning of the print, and we called that a disagreement between the score and the tape that the print would settle. This morning the 7-day score sits at 80, the 24-hour at 80, and the tape is moving toward the score rather than the other way around. First verdict: signal, not stubbornness. We are keeping the asterisk on that verdict anyway, because May earned it. Going into the May 20 print the tracked room was 43 to 7 bullish, the setup nearly identical, and Nvidia still gave back 10.2% over the following three weeks. One after-hours pop is not a grade. The grade comes from the same three-week window we held May to, which lands in mid-September, and we will run it in a weekly scoreboard whichever way it goes. ## Question three: the memory-stock rooms The [memory-stock piece](/blog/jim-cramer-memory-stocks) left a live test open: MU and SanDisk's tracked rooms stayed bullish through a selloff that kept deepening, and every thesis in those rooms runs through AI demand, which means it runs through Nvidia. Wednesday's print was the confirmation those rooms were waiting for, twice over: the $108 billion guide on the demand side, and on the pricing side Nvidia itself calling memory cost increases beyond its expectations, locking up capacity, and guiding its own margin down to absorb them. The memory makers' pricing power is now a line item in Nvidia's outlook. The tape had already started turning before the print: from Monday's closes to Wednesday's, Western Digital rose 7.7% to $468.88, Seagate 6.5% to $846.37, Micron 3.1% to $938.40, with SanDisk flat at $1,499.37. Overnight, Micron's room went 9 bullish to 0 with its tracked accounts at 3 to 0. The full ledger, though: from their August 17 closes, Micron is still down 7.3%, Western Digital 12.5%, and SanDisk 16.1%. The rooms held, the thesis got its confirmation, and the P&L has not caught up yet. Held conviction and paid conviction are different columns, and we will grade that one when the tape closes the gap, or refuses to. ## What this test was for One print settled the direction of three open questions, and it happened to settle them in the room's favor. Resist reading the win as the point. May was the same setup with the opposite result, and the next print is never far off. The part that survives either outcome is the timestamp: the crowd's split, the tracked accounts' lean, each name's track record, the score's disagreement with the tape, all on the record Tuesday morning while the print was still ahead. Thursday morning's 146 bullish mentions tell you what already happened. Tuesday's read told you who had conviction while conviction was still expensive. The same reading, updated live, is on the Nvidia page in Quantral now, next to the room's new debate about what $108 billion in guidance means for the rest of the AI complex. --- *Methodology: mention counts are subject mentions across the accounts Quantral tracks, bucketed in rolling 24-hour windows matching the in-app chart, bucket days August 6 through August 27; the final bucket covers the post-print reaction through early Thursday morning ET. Pre-print figures (386 mentions, the 3.2-to-1 crowd split, the 62-to-11 tracked split, May's numbers) are quoted as published in yesterday's piece. "Tracked" or "graded" means the author's credibility score is above 0.5 with at least ten lifetime graded calls. Prices are daily closes from our price feed; the after-hours move is as reported and may differ from today's session. 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.*