CoreWeave Launched Nvidia Vera Rubin NVL72 System
The cloud provider integrated the rack-scale AI platform to boost throughput for inference workloads.
Updated on Sept. 30, 2026 in Data Centers
Live Poll
Do you trust companies to securely integrate AI systems into critical clinical workloads?
CoreWeave has made the Nvidia Vera Rubin NVL72 system available on its cloud platform, offering a rack-scale design built to handle complex AI tasks. Early deployment shows the technology can significantly increase throughput for intensive inference workloads.
Why it matters
The system introduces specialized hardware, including the Vera CPU, which is engineered to accelerate agentic AI and resolve general-purpose infrastructure bottlenecks that limit performance.
The NVL72 rack-scale system integrates 36 CPUs and 72 GPUs, supported by 100 percent liquid cooling and cable-free modular tray designs. A single rack configuration supports up to 128 Vera CPUs and 11,264 cores.
The players
CoreWeave
This cloud infrastructure provider specializes in large-scale GPU resources for artificial intelligence and machine learning.
Nvidia
This technology company designs graphics processing units and data center systems critical for high-performance computing and AI.
Cognition
This AI research organization focuses on building autonomous systems and is an early adopter of advanced compute hardware.
Ennoble Care
This healthcare-focused firm entered a contract with CoreWeave to utilize Blackwell Server Edition nodes.
The details
The platform incorporates the NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, Spectrum-6 Ethernet Switch, Vera CPU, and Rubin GPU components. CoreWeave also introduced CoreWeave Forge, a development layer designed for building, improving, and evaluating AI models.
Timeline
Cognition began using the Vera Rubin system in early September 2026.
CoreWeave announced the platform availability on September 30, 2026.
Customers are expected to start testing standalone Vera CPUs in the coming weeks.
The Tech Race
This rollout accelerates the industry shift toward specialized, liquid-cooled, rack-scale computing designed specifically for agentic AI. It positions CoreWeave to compete by offering tighter integration of networking and compute silicon than legacy general-purpose data centers.
Companies and developers utilizing the platform can expect faster processing speeds for complex AI tasks, which may shorten development cycles for new software. These modular designs could lead to lower energy overheads as data centers transition to more efficient liquid cooling architectures.
The takeaway
The move to rack-scale, liquid-cooled systems signifies that AI infrastructure is moving away from modular, off-the-shelf components toward tightly integrated, purpose-built hardware. Organizations looking to scale agentic AI models will likely need to prioritize platforms that resolve these specific infrastructure bottlenecks.
What happens next
Customers are scheduled to begin testing the standalone Vera CPU bare-metal offering in the coming weeks.
Further reading
For more background on infrastructure shifts, visit the Data Centers section.
Live Poll
Do you trust companies to securely integrate AI systems into critical clinical workloads?
