CoreWeave Cloud Adds NVIDIA Blackwell RTX PRO 6000 GPUs

CoreWeave's Blackwell GPU Cloud: Power at Massive Scale

CoreWeave has expanded its cloud infrastructure offering by making NVIDIA RTX PRO 6000 Blackwell Server Edition GPU instances available at scale. The move allows customers to access Blackwell-based GPUs through a cloud model rather than through direct ownership of specialized hardware.

Invest in top private AI companies before IPO, via a Swiss platform:

By providing these GPUs as on-demand cloud resources, CoreWeave enables organizations to deploy high-performance AI and graphics workloads without purchasing dedicated servers or operating their own data center infrastructure. This approach can be particularly relevant for teams working on AI inference and multimodal applications that process combinations of text, images, audio, and video.

Blackwell Performance Revolution

The NVIDIA RTX PRO 6000 Blackwell Server Edition GPU is built on NVIDIA’s Blackwell architecture and delivers a measurable performance increase over previous generations. According to NVIDIA’s published specifications, the GPU can provide up to 5.6× higher throughput for large language model inference and up to 3.5× faster performance for text-to-video workloads compared with earlier architectures.

In practical terms, these improvements allow models to handle higher request volumes and deliver results with lower latency. This performance profile is designed to support models of up to approximately 70 billion parameters, a range commonly used in production environments where efficiency, responsiveness, and cost control are critical considerations.

By offering these GPUs through its cloud platform, CoreWeave allows customers to deploy models at this scale without managing hardware provisioning, cooling, networking, or ongoing maintenance.

Complete Blackwell Ecosystem Integration

The RTX PRO 6000 offering is integrated into CoreWeave’s broader Blackwell-based infrastructure portfolio. This portfolio includes large-scale systems such as NVIDIA GB200 NVL72, which is intended for extensive training workloads and high-density inference, as well as the NVIDIA HGX B200 platform, designed for modular high-performance training and inference deployments.

Within this lineup, RTX PRO 6000 instances occupy an intermediate tier. They are positioned for workloads that require high performance and efficiency but do not justify the operational complexity or cost associated with the largest multi-GPU clusters. This structure allows customers to select hardware configurations that align more closely with specific workload requirements.

Cloud Advantages Over On-Premise Solutions

Accessing Blackwell GPUs through the cloud provides operational flexibility compared with traditional on-premise deployments. Organizations can scale GPU capacity up or down in response to changing demand, rather than committing capital to hardware that may be underutilized outside peak usage periods.

CoreWeave manages hardware lifecycle operations, including upgrades and fault handling, which reduces the infrastructure burden on customer teams. As workloads evolve, users can transition between RTX PRO 6000 instances and larger Blackwell platforms within the same cloud environment, without undertaking new procurement or integration processes.

Democratizing Advanced AI Technology

Making Blackwell-based GPUs available through a cloud model broadens access to advanced AI compute resources that have historically been limited to organizations with significant capital and dedicated infrastructure. Developers, startups, and enterprises can now evaluate, test, and deploy AI and graphics workloads on the latest NVIDIA architecture using a consumption-based pricing model.

This approach lowers the threshold for experimenting with advanced AI capabilities and may shorten the time required to move from prototype to production. The availability of RTX PRO 6000 instances supports a range of use cases, including multimodal AI systems that combine language, visual, and video processing within a single deployment.