CoreWeave Bears Hold Ground as Analysts Pile In Bullish

The Street has been unusually active. JP Morgan's Samik Chatterjee upgraded the stock to Overweight from Neutral late last month, raising his target to $125. UBS and William Blair both initiated with positive ratings around the same time. Rosenblatt maintains a $250 target, the most optimistic on the board. The lone dissent came from Rothschild, which initiated at Sell with a $54 target. The mean analyst target across the coverage sits at $141, against a current price of $91.72, implying more than 50% return potential on the consensus view. The factor scores reflect the Street's enthusiasm: the analyst recommendation divergence score ranks in the 100th percentile of the universe, the highest possible reading. The short score of 63 and a short score rank in the 10th percentile tell the opposite story, marking CoreWeave as firmly in the most-watched-by-bears segment of the market. EPS momentum over 30 days sits in just the 6th percentile, suggesting near-term estimate revisions have softened even as longer-horizon growth expectations remain strong.

The bull-bear debate is unusually concrete. Bulls point to scarce NVIDIA GPU capacity, a $4 billion annualised revenue run rate, first-mover access to Vera Rubin NVL72 hardware with up to 4.8x throughput improvements, and a software stack through Forge and Serverless RL that extends monetisation beyond bare compute into higher-margin workflows. Bears lean on the concentration risk: roughly 67% of 2025 revenue came from Microsoft alone, with the top two customers accounting for 77% in 2024. EBITDA covers only 2.2x interest expense, and a 100 basis point rate move translates to roughly $107 million in additional interest cost. That is not a margin story; it is a financing risk story at scale.

Insider activity in the last week of September adds a layer of context. The Chief Strategy Officer, Chief Development Officer, and several other executives exercised options and sold into the market on September 30, with the CSO selling 65,616 shares at $87.69 for roughly $5.75 million and the CDO selling 56,531 shares for about $4.96 million. None of these sales were filed under a 10b5-1 pre-arranged plan. The net insider position over 90 days comes to minus $74.7 million across 859,000 shares. That is meaningful selling, though in context these are option exercises followed by partial sales, a common post-IPO pattern rather than a directional bet. The ORTEX 13D/G register shows no activist on the file. Magnetar Financial holds 14.9% as last disclosed, down from 23.3% previously, and NVIDIA Corporation holds 11.5%, up from 7.0%, a strategically meaningful position in its own right.

Retail attention has also been elevated. Wikipedia page views and ORTEX stock page traffic are running at a z-score of 1.74 above the company's own 90-day history, the highest attention reading in recent weeks. That is a measure of interest, not a revenue signal, but it tracks with the burst of analyst initiations and the post-earnings momentum. The previous Q2 print delivered a 22% one-day move; the next report is expected on November 12, giving the market 36 days to decide where to position.

With the stock up 7% on the week, short interest barely moving, analyst coverage expanding rapidly in the bullish direction, and borrow conditions that let bears sit comfortably, the key tension heading into November is whether the fundamental concentration and leverage concerns will draw in more shorts or whether the continued GPU demand narrative keeps them at bay.