CoreWeave, Inc. provides AI cloud infrastructure for compute-intensive training and inference workloads. Company news centers on GPU capacity, cloud platform performance, independent inference benchmarks, customer commitments, and product capabilities such as CoreWeave SUNK for deploying AI clusters across CoreWeave and multi-cloud environments.
Coverage also includes quarterly results, revenue backlog, capital-structure actions, private placements, senior notes offerings, and shareholder voting matters. Established in 2017 and listed on Nasdaq in March 2025, CoreWeave serves AI labs, startups, enterprises, and other customers building large-scale AI applications.
CoreWeave (CRWV) unveiled Physical AI Launchables, a catalog of tested workload templates for robotics, autonomous vehicle and simulation teams. Introduced at NVIDIA GTC Berlin, the catalog packages configurations built and tested by CoreWeave engineers, allowing teams to launch workloads with one command.
Launch templates cover NVIDIA Isaac Sim, CARLA, MuJoCo, NVIDIA AlpaSim and NVIDIA Cosmos 3. Each tool includes a blank template for custom workloads. The catalog also includes Cosmos Data Factory capabilities and agent skills for coding tools such as Claude Code and Codex. CoreWeave plans to add more Cosmos Data Factory workflows as templates.
CoreWeave (CRWV) launched its Zero Trust AI Framework, giving customers control over access to their models, data, and runtime environments. Unveiled at NVIDIA GTC Berlin, the framework combines six security layers covering encryption keys, provider access, infrastructure verification, protection during processing, model data, and network encryption. Customers can retain their existing key-management systems and authorize provider access for specific operational needs.
CoreWeave plans to extend NVIDIA confidential computing across CPUs, GPUs, and NVLink, beginning with select bare metal environments. HGX Rubin NVL8 is coming to CoreWeave Cloud in Q4 2026. The company also intends to support the NVIDIA OpenShell Reference Architecture.
CoreWeave (CRWV) announced its entry into India through an AdaniConneX collaboration, planning 240 megawatts of data center capacity in Navi Mumbai.
The planned deployment comprises three 80 MW buildings at AdaniConneX’s Taloja campus, with CoreWeave as the sole tenant. The company plans to deploy the NVIDIA Vera Rubin Platform and establish an Indian office with local hiring.
The first phase is expected to be available in mid-2028, with additional capacity coming online in phases. CoreWeave has the opportunity to expand by up to another 240 MW at the campus. The company said the expansion reflects growing customer demand for its AI Cloud. As of its August 11, 2026 earnings call, CoreWeave had approximately 4.2 gigawatts of contracted power.
CoreWeave (Nasdaq: CRWV) announced NVIDIA Vera Rubin NVL72 availability on its cloud, with Cognition already running production workloads on the system.
Cognition worked with CoreWeave to establish a cluster in early September. Its engineers measured 4.8 times total token throughput for SWE-2 inference workloads compared with a GB200 NVL72 baseline, and a 3.8 times increase in output token throughput for reinforcement learning workloads. CoreWeave also published measurements showing 10 times token throughput per megawatt versus NVIDIA GB200 NVL72 on the DeepSeek R1 reasoning model at matched interactivity.
CoreWeave (CRWV) has been selected by Ennoble Care to run clinical AI inference workloads across its home-based care network.
Under the agreement, Ennoble Care is deploying NVIDIA RTX PRO 6000 Blackwell Server Edition nodes on CoreWeave Kubernetes Service, a managed environment for running AI applications. The deployment provides dedicated, reserved GPU capacity and includes hardware-enforced isolation and zero-trust security controls. Ennoble Care provides home-based primary, palliative and hospice care across multiple states and is developing AI agents for clinical delivery and back-office functions.
CoreWeave (CRWV) launched CoreWeave Forge, a development platform connecting the work of building, running and improving AI models and agents. MasterClass and Canva are already building on the platform, which is available across models, frameworks and other clouds.
Forge combines Weights & Biases Models, OpenPipe post-training expertise and marimo-based notebooks with CoreWeave services. Its tools cover experiment tracking, evaluation, production-agent monitoring and versioning of model checkpoints and agent configurations. ARIA and Sandboxes are now generally available. Dedicated Inference is adding RL Rollouts in preview, allowing checkpoints to be loaded into a running deployment without redeploying. Forge offers Free, Pro and Enterprise editions, with a free 30-day trial for Pro.
CoreWeave (CRWV) launched the CoreWeave Partner Network, bringing technology integrations tested on its cloud under production conditions into one curated network. Available today, it brings together existing collaborations with partners including CrowdStrike, VAST Data, Reflection and ClickHouse.
The network covers infrastructure, data services, independent software vendors, and models and inference. CoreWeave plans to expand it to tools used by AI agents, starting with search. Exa, Parallel Web Systems and You.com are joining to provide a shared search layer through one integration rather than a separate contract for each.
CoreWeave (CRWV) announced plans to expand its computing portfolio with NVIDIA Vera CPUs to support workloads associated with AI agents. CPUs handle work surrounding model training and reasoning, including code execution and data pipelines. Vera will run on bare metal—directly on physical servers—under the same platform and consumption models as CoreWeave’s existing fleet.
CoreWeave Sandboxes provides hardware-isolated environments alongside training jobs. In testing, CoreWeave achieved more than 3x faster agent sandbox startup times on NVIDIA Vera CPU compared to an x86 CPU.
CoreWeave (CRWV) earned SemiAnalysis' Platinum ClusterMAX rating for a third consecutive time, remaining the only provider rated Platinum in every evaluation. ClusterMAX assesses cloud providers for large-scale AI workloads. In its latest evaluation, SemiAnalysis found that CoreWeave's health checks worked as intended and nearly all tests reached expected values out of the box.
CoreWeave was the sole Platinum-rated provider among roughly two dozen evaluated in the first ClusterMAX rating in early 2025. It retained that distinction in ClusterMAX 2.0 later that year, when the evaluation covered 84 providers. SemiAnalysis also uses CoreWeave's clusters as a reference point when discussing lower-rated providers.
CoreWeave (CRWV) announced a multi-year agreement under which Harell Data will deploy its research platform on CoreWeave Cloud to run AI model training, fine‑tuning and inference workloads on proprietary biotech datasets.
The Harell Data platform keeps raw scientific data inside its environment, bringing models to the data and preventing researchers from viewing or extracting underlying datasets. With CoreWeave’s NVIDIA A100 and NVIDIA Hopper GPUs, training runs are metered per job, enabling data owners to earn revenue share for each training task and allowing model builders to offer their models on a marketplace under a Models as a Service arrangement. Validation uses hidden test questions so model builders do not see answers and results are published against a single standard.