CoreWeave's Q2 FY2026 earnings press release carries a dateline of August 11, 2026, and is filed as an SEC exhibit (investors.coreweave.com, sec.gov). Q2 revenue was $2,575M, GAAP net loss was $626M, and revenue backlog reached about $104 billion, explicitly stated to exclude more than $25 billion of net new customer commitments added in early Q3 (same release). Adjusted EBITDA margin compressed year over year, 59% versus 62% (same release).

CRWV closed up 19.50% on August 12, 2026, the session after the release (tradingkey.com, low-tier aggregator, not cross-checked against a primary source). Today, August 13, is trading day T+2 off that release, and I found no CRWV-specific news item dated today itself. What is dated today is sell-side price-target revision reacting to Tuesday's print: Mizuho to $115 citing raised 2026 revenue guidance, Bernstein to $74 while keeping an Underperform rating (investing.com, investing.com), plus Argus to $150 and Cantor to $176 on the bullish side, and notably Barclays to only $105, a target that sits below today's tape, alongside Roth MKM trimming $150 to $145 (stockanalysis.com, medium-tier aggregator, notes not cross-checked against the original sell-side documents).

Intraday: open 107.25, spiked to 115.40 (+7.6% vs open) at 13:48Z, faded to 110.54 (+3.07%) by 15:15Z, so more than half the day's early gain has already been given back mid-session. Options flow dated Aug 12, 11:28am shows total volume over 436,000 contracts, calls 299,736 versus puts 136,396, heaviest single-strike open interest at the Aug 21 $100 call (61,088 contracts) and the $120 call (16,803 contracts), 3-month implied vol 80.9% (investing.com). That strike concentration brackets the range price traveled through today, a plausible dealer-hedging amplifier for the chop in both directions.

CRWV traded inside a same-day AI-infra sympathy cluster (SMCI, CLS, CIEN, NBIS, DOCN, CENX, BE), consistent with a shared "neocloud" earnings-week rally that also lifted Super Micro and Nebius, both of which reported the same week (finance.yahoo.com), while the broader tape was mostly red. Short interest was reported at 15.77% of float with 2.3 days to cover as of Aug 12, with that coverage framing the earnings-day pop as amplified by short covering (investing.com). CoreWeave's IPO lock-up itself expired a full year ago, August 2025, and insiders have sold over $2.3B since then as a described long-planned diversification, so lock-up expiry is not a live mechanism today (gurufocus.com).

Apply my three tests. No documented mechanical cause: no reconstitution, halt, forced liquidation, or fat finger is named anywhere; the lock-up is a year stale and the sourced short-covering language ties to Tuesday's pop, not to today specifically. No demonstrable misread: the bullish headlines (record backlog, beat and raise) track what the IR release itself says, and if anything the release understates the backlog by explicitly excluding $25B of new Q3 bookings, so there is no gap between headline and document to exploit. On magnitude, the one number I can actually compute cuts against calling this excessive: a ~$104B backlog (plus $25B+ more not even counted) sitting against a market cap around $61.0bn is a large figure pointing in the bullish direction, not the kind of "recall covers 3% of revenue but stock lost 19%" mismatch my worldview is built to catch. None of the three tests clears my bar, so by construction I stay off the overreaction side of the ledger and cannot score above 60.

The intraday shape reinforces staying low. The spike to +7.6% at 13:48Z and the fade to +3.0% by 15:15Z means the market has already priced out more than half of its own initial enthusiasm within the same session, before this is even measured. That lines up with documented short-term intraday reversal effects, where uninformed intraday flow gets a price concession from liquidity providers and reverses within the session or the next few days, an effect the literature finds is strongest in illiquid, high-beta names like this one (quantpedia.com). What remains at the +3% mark reads as ordinary two-session digestion of Tuesday's real, primary-sourced beat and raise, reinforced by a genuine batch of same-day sell-side target increases (medium tier, second-hand, but numerous and directionally consistent).

I land at justified, not the floor of the scale, because Barclays is on record today with a Hold and a $105 target sitting below the current tape (stockanalysis.com), a genuine, sourced dissent that the stock may already be running ahead of fundamentals. That keeps my score off the true floor and my confidence capped, since the specific catalyst for today's incremental move is not confirmed by any primary document. I am not making an explicit, sourced momentum case strong enough to justify a directional call, so direction is none and expected_reversion_pct is flat.

The strongest argument against me is Barclays' own number: a Hold rating with a $105 target below spot is a sell-side voice saying, today, that the market has already overpriced this. If that view is right, the day's continuation is an overreaction that should mean-revert lower, which would flip both direction and score. I could not corroborate that note against its original text, only a secondhand aggregator, so I am naming it as a caution rather than leaning on it as proof.

The options open-interest data itself is real (fetched directly from investing.com), but a hostile read could not confirm the specific put/call OI ratio quoted alongside it, so I dropped that one figure. Without a clean ratio I can describe the $100/$120 strike concentration as a plausible dealer-hedging amplifier for today's chop, but I cannot say with confidence which direction dealer gamma pushes price from here.

There is an unresolved contradiction on the earnings beat itself: a hostile read of one source's own text reported an adjusted EPS loss of $1.14 against a $1.41 consensus, materially different from the $1.03-versus-$1.20 figure that did not survive verification elsewhere. I do not have a clean, confirmed EPS-beat number I trust. That is a real hole under the "real beat" half of my case, though not fatal since revenue and backlog are independently confirmed by the primary release.

Taking the underreaction question seriously, as my own worldview requires: if the unverified +246% YoY backlog growth figure and the unverified FY26 guidance raise turn out accurate once a working copy of that release can be read, the fundamental case is stronger than what I could confirm here, arguing the market is still underpricing this name rather than overpricing it, which is also inconsistent with an overreaction call and would only reinforce staying at justified or lower.

There is also a same-name precedent for gains not holding: after CoreWeave's Q1 2026 report, the stock round-tripped down 33.4% over the following three months amid a securities-fraud suit, CEO stock sales, and a competitive announcement from Meta (247wallst.com), though that slide had multiple compounding negative catalysts of its own and unfolded over months, not the next session, so it is a weak analogue for tomorrow specifically. An unverified Benzinga headline separately claims CoreWeave "averages a 17% drop" after its last four earnings beats; I could not open the underlying article (403), so I am not treating it as evidence, only flagging it as an unresolved pattern claim worth checking before leaning too hard on justified.