Ten days after Rothschild's Sell call, someone paid $2.3M for CRWV puts that stop right above the 52-week low
CRWV trade card · OptionWhales daily thesis
Wall Street still carries CoreWeave at a consensus Buy, 35 analysts averaging a price target near $138.78 per MarketBeat, while the stock changed hands at $88.33 on Thursday, October 1, 2026, barely half its 52-week high of $153.20. A trader leaned the other way at 13:50:59 ET, paying a net $2.3M for a 5,000-contract CRWV put spread expiring January 15, 2027: long the 90s, short the 70s, one print. The structure is bearish. Both legs were marked at nearly identical implied vol, so the vega offsets and what the debit buys is direction.
The bear case here got a name ten days ago. Rothschild Redburn initiated CoreWeave at Sell on September 21, 2026, arguing that falling GPU-rental prices, hyperscalers building their own capacity and expensive financing would compress returns (Yahoo Finance, September 23: https://finance.yahoo.com/technology/ai/articles/coreweave-nebius-just-got-sell-185913300.html). Invezz on September 29 had the stock in a bear market with short interest above 10%, a figure about the shares generally and not about this order. Under that argument sit $2.58B of Q2 revenue against a $626M net loss, roughly $33.8B of total debt as of June 30, capex guided to $35B-$39B for 2026, and a $104B backlog. The day's actual headline pointed elsewhere: Zacks, October 1, on the new CoreWeave Partner Network and the NVIDIA Vera CPU deployments unveiled at Fully Connected the day before, a story about ecosystem reach, which is a separate question from what a GPU-hour rents for.
The two legs:
- bought 2,500 of the Jan 15, 2027 90 puts for $3.6M, $14.40 a share, delta -0.43, IV 74.7%
- sold 2,500 of the Jan 15, 2027 70 puts for $1.3M, $5.20 a share, delta -0.21, IV 75.0%
Net $9.20 a share for a $20-wide window, 106 days out, with the long strike $1.67 above spot and the short strike about 21% below it. Net delta works out near -0.22 per spread. Prior-day open interest was 6,791 at the 90s and 33,188 at the 70s, both far larger than the 2,500 traded in each leg, so either side could have been opened or closed inside that interest and open versus close is not determinable here.
What interests me is where they stopped. Selling the 70s caps the structure above the 52-week low of $60.55, which means the thing being bought sits in the band between today's price and the bottom of the existing range, and the tail underneath it was sold off to pay for the near part. Anyone pricing the financing accident implied by $33.8B of debt and compressing rental economics would want to own that tail, and this structure sells it. Paying 46% of the width for a window anchored at the money, at 75 vol, is a workable way to own a move into the 70s and a poor way to own anything under them. A second reading stays live: with short interest where it is and the equity sitting on top of that debt load, this could be financed protection on exposure I cannot see. The strike placement persuades me more than the hedge reading does, because a hedger with that much balance-sheet risk in mind usually reaches further down, not to a strike already in the money.
CoreWeave's next earnings report is estimated for November 9, 2026, inside the contract's life, and the expanded $21B Meta agreement announced April 9, 2026 covers capacity running 2027 through 2032, so this spread expires two weeks into the first year of that contract. If GPU-hour pricing keeps slipping while interest expense compounds, the 90-70 band is exactly where this position lives. What breaks it is utilization and contract pricing holding up, and November is the first dated read on which of those is happening.
*Education, not advice. Flow data from optionwhales.io.*