NVIDIA Vera Deployment
CoreWeave deploys NVIDIA Vera Rubin systems to accelerate agentic AI workloads.
Anonymous
Key Points
1
CoreWeave said it raised prices by 25% in July and by a further 10% since then; analyst Brett Knoblauch expected the increases to begin affecting results in 2027 and more substantially in 2028.
2
CoreWeave cited Ennoble Care, Canva and Capital One as new enterprise inference clients, diversifying its customer base beyond AI labs.
3
CoreWeave said Cognition began using Vera Rubin systems in early September. CoreWeave executive Chen Goldberg said the company got production workloads running within days, attributing the rapid deployment to years of engineering its platform across GPU generations.
4
CoreWeave described agentic AI workloads as involving bursts of CPU-intensive tasks—including isolated sandboxes, tool calls, code execution and data pipelines—that may require thousands of environments for a short period before demand drops. The company said Vera runs on its cloud under the same platform and consumption models as the rest of its fleet.
CoreWeave is now offering NVIDIA's Vera CPU and has begun deploying Vera Rubin NVL72 systems for production customers. The company's Vera rack configuration delivers 128 CPUs and 11,264 cores, supporting more than 11,000 concurrent agent environments with sandbox startup times more than three times faster than previous setups.
Cognition, the first production customer for the NVL72 systems, reported up to 4.8 times the token throughput on inference workloads and 3.8 times the output-token throughput on reinforcement-learning tasks compared to GB200 NVL72 baselines. An analyst maintained an Overweight rating with a $176 price target, citing strong demand and potential benefits from CoreWeave's recent price increases.
Vera Powers AI Agent Workloads at Scale
CoreWeave designed the Vera rack specifically for agentic AI. These workloads involve rapid bursts of CPU-intensive tasks—isolated sandboxes, tool calls, code execution, and data pipelines. Companies like Cognition may need thousands of environments for just hours before demand drops, then scale back.
CoreWeave executive Chen Goldberg said production workloads went live within days. He credited years of engineering the platform across GPU generations. Vera runs on CoreWeave's cloud under the same platform and consumption models as its existing fleet, simplifying adoption.
Enterprise Clients Drive Demand Beyond AI Labs
CoreWeave is no longer dependent on AI research labs. New enterprise inference clients include Ennoble Care, Canva, and Capital One, diversifying revenue streams. The company reported demand for capacity substantially exceeds supply and reiterated confidence in its 2030 capacity expansion goals.
Price Increases Begin Flowing Into 2027 Results
CoreWeave raised prices by 25% in July and another 10% since then. Analyst Brett Knoblauch expects these increases to begin affecting quarterly results in 2027 and more substantially in 2028. Higher pricing combined with enterprise growth could drive margin expansion.
New Platform Forge Supports End-to-End AI Workflows
CoreWeave launched Forge at its Fully Connected conference, a platform for model training, evaluation, and deployment. The company also announced a partner network with integrations including CrowdStrike, VAST Data, Reflection, and ClickHouse, expanding ecosystem support for complex AI infrastructure.
Sentiment
Mostly PositiveNEGATIVENEUTRALPOSITIVE
Publishers
15
Articles
18
Reach
33
A
AnonymousAuthor
