Nebius stock jumped 34%, CoreWeave 19% - can it last?
13 August 2026 · 12 min read
The short version: Nebius closed 34.14% higher on 12 August and CoreWeave 19.28% higher - the two largest moves on our 34-name board, both on second-quarter results. Nebius drew 9,116 posts on X that day, the largest single-day count any ticker on this board has recorded since we started counting, and more than NVIDIA has ever managed in a day.
Our market sentiment score puts Nebius at 86 and CoreWeave at 82, third and sixth of 34. Here is the part most coverage skips: those scores are almost entirely out of room. Four of the six components that build them sit within 0.15 of a point of their arithmetic ceilings. On our own arithmetic, a single flat session would take Nebius from 86 to 67 without one other input changing.
All board figures are as at the close on 12 August 2026, the last complete session. This piece is about the two names; the segment-wide split across all six neoclouds is a separate measurement.
What happened to NBIS and CRWV stock on 12 August?
Two things about that session put these moves in context. Only five of the 34 names moved more than 10% on 12 August, and Nebius and CoreWeave were the two largest of them. Only four names traded at more than twice their own normal volume, and both of these were among them. This was not a broad melt-up carrying everything along: the board's median score that session was 66, a full 20 points below Nebius.
Nebius is now the more recovered of the two. It sits 13.6% below its 52-week high of $299.86, so it needs 15.7% to reclaim it. CoreWeave, despite the same kind of session, is still 29.7% below its high of $153.20 and would need 42.2% to get back. Both had fallen a long way first: CoreWeave traded at $60.82 on 29 July, twelve sessions before this.
Why is Nebius stock up? The Q2 numbers
Nebius reported before the open on 12 August, and the growth rate is the headline that did the work:
- Revenue of $582.3 million, up 454% year on year from $105.1 million. First-half revenue of $981.3 million was up 529%.
- Adjusted EBITDA of $236.2 million, roughly a 41% margin, against a loss in the same quarter of 2025. That is the swing that matters: this is a business that stopped burning at the operating line while growing more than fivefold.
- Adjusted net loss of $33.2 million, narrowed from $91.5 million a year earlier.
- Annualised run-rate revenue of $3.0 billion at the end of June, up 56% on the first quarter, helped by four AI cloud contracts each averaging over $1 billion of total contract value.
- Capital expenditure of $5.657 billion in the quarter and $8.130 billion across the half, against $8.499 billion of non-current debt at 30 June.
- Management raised its year-end contracted power target to 5 GW and reaffirmed full-year guidance.
Behind the quarter sits the contracted book the company disclosed in July: more than $40 billion of additional contracted revenue from investment-grade customers including Microsoft and Meta. Against a $65.2 billion market capitalisation, that is roughly 0.6 times - a very different ratio from the bitcoin-mining converters, where contracted revenue runs to multiples of market value.
Why is CoreWeave stock up? The Q2 numbers
CoreWeave reported after the close on 11 August, which is why its attention spike lands a day before its price spike. The scale is an order of magnitude above Nebius:
- Revenue of $2.575 billion, up 112% from $1.212 billion, and up 24% on the first quarter.
- Revenue backlog of approximately $104 billion at 30 June, up 246% year on year, with more than $25 billion of net new customer commitments added in early Q3 on top.
- Net loss of $626 million, widened from $290 million. Adjusted operating income of $128 million on a 5% margin, against $200 million and 16% a year earlier - revenue doubled and adjusted operating income fell.
- Active power expanded by nearly 500 MW to 1.5 GW, with total contracted power at approximately 3.7 GW. Fifty-one active data centres at 30 June.
- Full-year 2026 guidance raised to $12.4-13.2 billion of revenue and $960 million to $1.15 billion of adjusted operating income, with third-quarter revenue guided to $3.45-3.6 billion.
- New customers named in the quarter included Bentley Systems, Caterpillar, Grammarly, Isomorphic Labs and Sunday Robotics.
That $104 billion backlog is 1.8 times CoreWeave's entire market capitalisation, and 2.2 times if the early-Q3 commitments are included. The bear case is not that the demand is fake. It is that the margin went the wrong way while capital expenditure and debt went up, and that a backlog is only worth what it costs to deliver.
The market is paying five times as much for one as the other
Here is a comparison we have not seen made, and it is straightforward arithmetic on the two companies' own disclosures.
CoreWeave guided third-quarter revenue to a $3.525 billion midpoint, an annualised run-rate of about $14.1 billion. Nebius ended June at a $3.0 billion annualised run-rate. Set those against market capitalisations of $58.8 billion and $65.2 billion:
CoreWeave is running at roughly 4.7 times Nebius's revenue for a smaller market capitalisation. Per dollar of run-rate revenue, the market is paying about five times as much for Nebius.
That gap is not irrational, and it is worth saying why rather than presenting it as a mispricing. Nebius is growing four times faster off a much smaller base, and it converted to positive adjusted EBITDA at a 41% margin while CoreWeave's adjusted operating margin compressed from 16% to 5%. One is being priced on the slope, the other on the level. The multiple is the market's estimate of how long the slope holds - and that is a forecast, not a measurement, which is precisely why our score does not attempt it.
The attention spike is the largest we have ever recorded
We count posts on X carrying each ticker's cashtag ourselves, one row per complete New York day, retweets excluded. On 12 August the whole 34-name board carried 32,161 posts. The distribution was extraordinary:
Three findings from that table are worth stating plainly.
Nebius's 9,116 posts is the highest single-day count for any ticker on this board since the series began on 3 August. The previous high was NVIDIA's 8,104 on 4 August. A $65 billion company out-posted the most-discussed $5.4 trillion company's own record.
Both names out-posted NVIDIA on the day. NVIDIA is 83 times larger than Nebius by market value and drew 40% as many posts. Measured per billion of market capitalisation, Nebius drew 200 times NVIDIA's attention.
Between them the two took 40.0% of the entire board's conversation. Two names out of 34 took two-fifths of the day. And for Nebius the arrival was almost entirely on that one day: its 9,116 posts are 37.5% of everything it has accumulated across the ten days we have counted.
Posts on X carrying each ticker's cashtag, counted by this site over 27 August 2026. Retweets excluded. See all 34 names →
One thing this number is not: a direction. We measured this on our own data and published the result - across all 34 names on a full trading day, the rank correlation between post volume and the day's price move was -0.00. Post counts track dollar turnover, not direction. A collapse and a rally generate the same kind of number. The 9,116 tells you where the money and the eyes went; it says nothing whatsoever about which way the stock goes next.
What our sentiment score says - and what it does not
Nebius at 86 is one of only three names on the board in the Overheating band, alongside Lumentum at 87 and Arista at 86. CoreWeave at 82 is the third-highest of the ten names reading Hot.
The score is built from six price-and-volume components, each weighted and each compressed through a function that approaches its bound without reaching it. Here is what actually produced those two numbers, with each component's ceiling in brackets:
Read the top four rows. Nebius has 38.25 points of a possible 38.40 from momentum, relative strength, trend and volume - 99.6% of the available total. CoreWeave has 38.36, or 99.9%. There is essentially nothing left to win there.
The only headroom is in the two components that are currently negative for both names. Short-run volatility is running 22.0% above Nebius's own longer-run norm and 21.1% above CoreWeave's, which is exactly what a 34% day does. To score higher, these two would have to keep rising while becoming calmer and while closing the gap to their 52-week highs. That is not impossible. It is the opposite of what has been happening.
Can the run continue? Five indicators, honestly read
The first thing to say is the thing a site built on this number has to say: the market sentiment score is a description of how a stock is behaving, not a forecast of what it will do. It is recomputable from public price data by anyone, which is the single most valuable property it has, and it is a property you only keep by refusing to dress it up as prediction. With that stated, here is what the indicators show.
1. Roughly 40% of the score turns on a single session. Momentum is 25% of the weight and the volume component is signed by the day's direction, adding another 15%. So an 86 recorded the day after a +34.14% session is substantially a restatement of that session. Holding every other input at its 12 August value - a simplification, since the five-day and trend components would move too - the mechanical effect of the next session alone is this:
A flat day costs Nebius 19 points. A 3% decline costs it 36 and takes it to the middle of the board. This is not a criticism of the score - it is the score doing what it is designed to do, which is describe today rather than guess at tomorrow. But it means an 86 should be read as "this stock had an exceptional session on exceptional volume", not as "this stock is set fair".
2. Our own record for these two names is that high scores have not persisted. We can only show 14 sessions here, because 30 days of sentiment history is what the public API serves, but within that window the pattern is stark. Nebius scored 77 on 3 August. Three sessions later it scored 15, with the price 10.7% lower. Its 75 on 4 August became 24 by 7 August, price 16.7% lower.
Across those 14 sessions Nebius's score ranged from 5 to 86 and CoreWeave's from 9 to 82, with a mean absolute day-to-day change of 18.9 and 18.8 points respectively. These are two of the most volatile readings on the board. Four episodes is an anecdote, not a study, and we will not pretend otherwise - but it is our own anecdote, on our own numbers, and it points one way.
3. Volatility is the one component pushing back, and it is doing so on both. The negative volatility readings above are not noise. They are the formula recognising that short-run price dispersion has run well ahead of each name's own normal. That component penalises violent rallies exactly as it penalises selloffs, which is deliberate: the question it asks is whether uncertainty is increasing, and after a 34% day the answer is yes.
4. The professionals disagree by more than two to one. On 12 August, analysts revising CoreWeave targets after the same set of numbers landed on $155 (Truist, Buy), $153 (Piper Sandler), $115 (Mizuho, Neutral) and $74 (Bernstein, Underperform). Bernstein raised its target while keeping an Underperform rating. A 2.1x spread between the highest and lowest published target on the same day, on the same disclosures, is a reasonable measure of how much of this is estimation rather than measurement.
5. The gap to the highs is very different for the two. Nebius needs 15.7% to reclaim its 52-week high; CoreWeave needs 42.2%. If the argument is that these are recovery trades, they are at very different points in that recovery, and CoreWeave's drawdown component is still costing it 3.97 points where Nebius's costs only 0.50.
What we are not claiming
Three things, stated so they cannot be read into the above.
We are not saying the news caused the move. Both companies reported and both stocks rose sharply, and the sequence is clear enough. But most financial coverage is written after a move and about it. Coverage is evidence of what was published, not proof of causation, and we apply that rule to our own writing as much as anyone else's.
We are not saying the mention counts predict anything. We have been counting cashtags since 3 August. Ten days is not a sample. Testing whether today's posting says anything about tomorrow's return needs enough history that a run of luck cannot carry the result, and when we have it we will publish the answer whichever way it falls.
We are not folding attention into the score. Nebius taking 28.3% of the board's conversation is a fact we can prove and it changes the sentiment score by exactly zero. Crowd measures are displayed on this site and never scored, because the moment an opinion goes into the formula it stops being reproducible from public price data by anybody who wants to check it.
The short side, and the limits of what we know
Michael Burry of Scion Asset Management disclosed a short position in Nebius on 7 August, at a reported entry of $211.77, with a thesis that GPU depreciation runs faster than neocloud operators model it. Our board's close on 12 August was $259.20, which is 22.4% above that reported entry.
The limits of that are worth stating as plainly as the number. The entry price is as reported rather than independently verified by us; we do not know the size of the position, what instruments it uses, or whether it is still open. A position being underwater five sessions in says nothing about whether the depreciation argument is right - that is a question about accounting over years, and it will be settled by cash flows rather than by a fortnight of price action.
What to watch next
- The first session that is not up. On the table above, a single flat day costs Nebius 19 points. The score's behaviour on the next ordinary session is the cleanest read on how much of the 86 was the event and how much is a state.
- Whether the attention holds. Nebius's 9,116 posts were 37.5% of its entire ten-day total. Event spikes decay fast. If the count settles back near its 1,000-2,000 baseline within a week, the interest was the earnings; if it holds above 4,000, something has changed about who follows the name. The X mentions board updates daily.
- CoreWeave's Q3 margin. Revenue doubled and adjusted operating income fell. Guidance implies $960 million to $1.15 billion of adjusted operating income for the full year against $128 million in Q2, so the back half has to do most of it. That is falsifiable in November.
- Nebius's capacity coming online. The company delivered $582.3 million of revenue and a $3.0 billion run-rate before most of its 2026 capacity is live, and raised its year-end contracted power target to 5 GW. The conversion of contracted power into revenue is the whole thesis.
- Financing. Nebius spent $8.130 billion of capex in the half against $8.499 billion of non-current debt; CoreWeave's capital programme is larger again. Dilution and interest are the mechanism by which a large backlog fails to reach shareholders, and it applies to every name in this segment.
The neocloud segment page carries the live version of every score quoted here, and the X mentions board carries the current post counts.
Frequently asked questions
Why is Nebius (NBIS) stock up?
Nebius reported second-quarter 2026 results before the open on 12 August 2026 and closed 34.14% higher at $259.20, the largest move on the ChipSentiment board that session. Revenue was $582.3 million, up 454% year on year, with adjusted EBITDA of $236.2 million at roughly a 41% margin against a loss a year earlier, an adjusted net loss narrowed to $33.2 million, and annualised run-rate revenue of $3.0 billion at the end of June, up 56% on the first quarter. Management also raised its year-end contracted power target to 5 GW.
Why is CoreWeave (CRWV) stock up?
CoreWeave reported second-quarter 2026 results after the close on 11 August 2026 and closed 19.28% higher at $107.73 on 12 August. Revenue was $2.575 billion, up 112% year on year, with a revenue backlog of approximately $104 billion at 30 June, up 246%, plus more than $25 billion of net new customer commitments added in early Q3. Full-year 2026 revenue guidance was raised to $12.4-13.2 billion. The offsetting figure is margin: adjusted operating income was $128 million on a 5% margin, against $200 million and 16% a year earlier.
What is the ChipSentiment score for Nebius and CoreWeave?
As at the close on 12 August 2026, Nebius scored 86 out of 100 and CoreWeave 82, ranking third and sixth of the 34 names tracked, against a board median of 66. Nebius is one of only three names in the Overheating band. The score is built from six price and volume components and contains no news or social input, so it can be reconstructed by anyone from public price data.
Are NBIS and CRWV expected to keep rising?
The ChipSentiment score describes how a stock is behaving rather than forecasting what it will do, so it cannot answer that directly. What it can show is that both scores are near their arithmetic limit: four of the six components sit at 99.6% (Nebius) and 99.9% (CoreWeave) of their combined ceiling, and the only remaining headroom is in volatility and drawdown, both of which are currently negative. Holding all other inputs constant, a single flat session would take Nebius from 86 to 67 and CoreWeave from 82 to 63. In the 14 sessions of history available, Nebius scored 77 on 3 August and 15 three sessions later.
How many X posts did Nebius get after earnings?
Nebius drew 9,116 posts carrying its cashtag on 12 August 2026, up 277% from 2,416 the previous day. That is the highest single-day count recorded for any of the 34 tickers on the board since counting began on 3 August, exceeding NVIDIA's previous high of 8,104. It represented 28.3% of the board's 32,161 posts that day, and Nebius and CoreWeave together took 40.0%. Post counts measure attention, not direction: across the board on a full trading day the rank correlation between post volume and the day's price move was -0.00.
Which is more expensive, Nebius or CoreWeave?
On run-rate revenue, Nebius. CoreWeave guided third-quarter revenue to a $3.525 billion midpoint, an annualised run-rate of about $14.1 billion, against a $58.8 billion market capitalisation - roughly 4.2 times. Nebius ended June at a $3.0 billion annualised run-rate against a $65.2 billion market capitalisation, roughly 21.7 times. The market is paying about five times as much per dollar of run-rate revenue for Nebius, which grew 454% against CoreWeave's 112% and posted a positive 41% adjusted EBITDA margin while CoreWeave's adjusted operating margin fell from 16% to 5%.
Sources
- Nebius Group - second quarter 2026 financial results - revenue $582.3m (+454% year on year from $105.1m), H1 revenue $981.3m (+529%), adjusted EBITDA $236.2m at approximately 41% of revenues, adjusted net loss $(33.2)m against $(91.5)m, Q2 capex $(5.657)bn and H1 capex $(8.130)bn, non-current debt $8.499bn at 30 June 2026
- 24/7 Wall St - Nebius Group Q2 2026 earnings - $3.0bn annualised run-rate at end of Q2, up 56% on Q1; four AI cloud contracts averaging over $1bn of total contract value each; year-end contracted power target raised to 5 GW; full-year guidance reaffirmed
- Nebius Group - Form 6-K, secured debt facility (SEC EDGAR) - more than $40bn of additional contracted revenue from investment-grade customers such as Microsoft and Meta
- CoreWeave - second quarter 2026 results - revenue $2,575m against $1,212m (+112%), revenue backlog approximately $104bn at 30 June with more than $25bn of net new commitments in early Q3, net loss $(626)m against $(290)m, adjusted operating income $128m at 5% margin against $200m at 16%, active power expanded by nearly 500 MW to 1.5 GW and total contracted power approximately 3.7 GW, new customers including Bentley Systems, Caterpillar, Grammarly, Isomorphic Labs and Sunday Robotics
- Investing.com - CoreWeave Q2 2026 slides - full-year 2026 revenue guidance $12.4bn-$13.2bn, adjusted operating income $960m-$1.15bn, Q3 revenue guidance $3.45bn-$3.6bn, 51 active data centres at 30 June
- Investing.com - Mizuho raises CoreWeave price target - target raised to $115 from $100, Neutral rating maintained
- Investing.com - Bernstein raises CoreWeave price target - target raised to $74 on the revenue beat and raised fiscal 2026 guidance, Underperform rating maintained
- Finbold - Wall Street updates CoreWeave price targets after double earnings beat - Truist raised to $155 from $126 with a Buy rating; Piper Sandler raised to $153
- 24/7 Wall St - Michael Burry is short Nebius and Palantir - short position in Nebius disclosed on 7 August 2026 at a reported entry of $211.77, on a thesis that GPU depreciation runs faster than neocloud operators model it; position size and current status not disclosed
- ChipSentiment - X mentions board - first-party cashtag post counts, one row per complete New York day, retweets excluded: NBIS 9,116 and CRWV 3,745 on 12 August 2026 against NVDA 3,702, board total 32,161
- ChipSentiment - how the market sentiment score works - six price and volume components and their weights, and the reason no news or crowd input enters the formula
Figures are taken from the public filings and the reporting linked above.