Overview
- CoreWeave said it has validated and deployed Nvidia’s Vera Rubin NVL72 at multi‑rack scale, becoming the first AI cloud provider to do so and also rolling out AI Object Storage features such as cross‑region write acceleration, an Archive tier, and a Local Object Transport Accelerator that the company says cuts latency and boosts per‑GPU throughput.
- The company first filed to sell $3.0 billion of convertible senior notes in a private placement and an ATM for up to 35 million Class A shares on Thursday, and then priced and upsized the convertible offering to about $3.7 billion on Friday with an initial conversion price near $97.85 per share.
- Markets reacted in two phases: shares rose roughly 3% on the NVL72 deployment, then fell about 4% after the financing and ATM were disclosed, and Cathie Wood’s ARK bought roughly 239,083 shares during the dip while filings show heavy insider selling in recent months.
- CoreWeave reported strong top‑line momentum — Q2 revenue of $2.58 billion, a large contracted backlog and about 4.2 GW of contracted power — but it remains unprofitable, carries large existing debt and high interest costs, and needs repeated large capital inflows to fund data‑center buildout.
- Key near‑term watch items are the final pricing and execution of the convertibles and the capped‑call hedges that are meant to limit dilution, how counterparties’ hedging or stock purchases move the share price, and whether sustained customer demand will cover rising interest and capex burdens.