CoreWeave

Market position

CoreWeave describes itself as an AI hyperscaler and was recognized by Gartner as a Visionary in the 2026 Magic Quadrant for Cloud AI Infrastructure. It is a Nasdaq-100 component with a market capitalization near $49.5 billion as of August 2026, and its 2025 IPO was the largest AI-related listing by amount raised according to Dealogic. It was reported to be the first cloud provider to offer Nvidia GB200 NVL72 chips and the first to commercially deploy GB300 NVL72 GPUs. Yahoo Finance groups it in the Software—Infrastructure industry alongside companies such as Oracle, Palantir, CrowdStrike, Microsoft and Palo Alto Networks, and press coverage compares it with other AI-cloud stocks including Nebius and IREN.

CoreWeave differentiates on purpose-built, Kubernetes-native AI infrastructure rather than general-purpose cloud, early and privileged access to new Nvidia GPU generations, bare-metal performance with automated provisioning, cluster health and lifecycle management, and claimed operational metrics including 10x faster inference spin-up, up to 96% goodput and 50% fewer daily interruptions. It also cites transparent pricing with no hidden fees, open standards, and dedicated engineering support. Customer references cite outcomes such as 2.5x faster training for Mistral AI, up to 80% faster AI workloads for IBM, and 3x faster request serving with 75% cloud cost savings for NovelAI.

Technology

The platform is Kubernetes-native and built on bare-metal infrastructure with automated provisioning and support for leading workload orchestration frameworks. Components include the CoreWeave Kubernetes Service, Fleet and Node LifeCycle Controllers, Mission Control lifecycle management, the Tensorizer runtime acceleration tool, a Local Object Transport Accelerator for data and storage, observability and security tooling, and the Weights & Biases AI developer platform (including W&B Weave for tracing and monitoring agents). CoreWeave also develops its own chip management software and is an early deployer of new Nvidia hardware generations, including H100, Blackwell, GB200 NVL72, GB300 NVL72 (Blackwell Ultra) and Vera Rubin NVL72 systems. The company reports record-setting MLPerf results and cites up to 96% cluster goodput.

Go-to-market

CoreWeave sells directly to large AI labs and enterprises through named contracts and multi-year capacity agreements, supported by a senior commercial organization (Chief Revenue Officer, EVP Strategic Deals) and 24/7 direct-to-expert engineering support with hands-on onboarding. Marketing assets include customer case studies, published pricing, a TCO calculator, MLPerf benchmark results, webinars and events, and a startup accelerator program launched in October 2022 that provides compute credits, discounts and hardware resources.

AI labs, cloud platforms, enterprises and quantitative trading firms running large-scale training and inference workloads, plus users of visual effects and rendering, pixel streaming and batch processing. Named customers and references include OpenAI, Mistral AI, IBM, Jane Street, NovelAI, Odyssey, MasterClass, Stability AI, Hudson River Trading, IMC, Anthropic and Meta Platforms.

Geography

Headquartered in Livingston, New Jersey (earlier releases cite Roseland, N.J.), with data centers operating in the United States and Europe. In 2023 the company expanded from three to 14 data centers and expected 28 globally by the end of 2024. It established a European headquarters in London in 2024 with a $1.25 billion regional investment commitment and announced plans to invest $978.6 million in UK data centers. It also built a $1.6 billion supercomputer data center for Nvidia in Plano, Texas. Wikipedia lists its area served as international.

History

The company was founded on September 21, 2017 in New Jersey as Atlantic Crypto, initially mining ethereum with GPUs. Following the 2018 cryptocurrency crash, it was renamed CoreWeave in 2019 (the name change to CoreWeave, Inc. is dated December 2019 in one source) and repurposed its GPU inventory into cloud computing infrastructure. It ran three data centers in 2022, quadrupled that to 14 during 2023, and targeted 28 globally by the end of 2024; headcount grew from three founders to about 550 people by May 2024. In summer 2022 it spent roughly $100 million on Nvidia H100 chips, and its privileged access to Nvidia GPUs drove rapid growth as AI demand rose in 2022-2023, with customers such as Stability AI. It launched a startup accelerator program in October 2022. Nvidia invested $100 million in April 2023, and the company was valued at about $2 billion by May 2023. CoreWeave filed its Form S-1 in March 2025 and listed on Nasdaq as CRWV on March 28, 2025.

Risks & controversies

The company remains unprofitable, reporting a 2025 net loss of $1.2 billion and a trailing-twelve-month net loss of $1.93 billion, with a total debt-to-equity ratio of about 1,027% and levered free cash flow of -$9.09 billion; Yahoo Finance coverage cites profitability challenges, roughly $640 million of interest expense, and financing concerns weighing on AI-cloud stocks. Customer concentration is high, with Microsoft representing over 60 percent of 2024 revenue per the S-1. Acquisition attempts for Core Scientific failed twice: a $1 billion offer was rejected in June 2024 and a $9 billion agreed deal was voted down by Core Scientific shareholders in October 2025. Its use of Nvidia H100 GPUs as debt collateral in 2023 drew notice in the business press, and the company reduced its IPO size from $2.7 billion to $1.5 billion shortly before listing.