CoreWeave tests Nvidia’s Rubin NVL72 in AI cloud milestone

The move comes as enterprises look for ways to deploy AI models at scale, with inference workloads becoming a key focus for cloud providers.

SiliconANGLE reported the validation, noting that CoreWeave’s infrastructure is built for AI from the start, unlike general-purpose clouds that adapt existing data centers. The Rubin NVL72 is part of Nvidia’s next wave of AI hardware, and early testing by a cloud provider suggests a push to support newer chips faster. CoreWeave hasn’t shared details on pricing or when Rubin-based services might launch, but the timing could give it an edge if demand for inference-optimized infrastructure keeps rising.

This fits into a larger shift: AI workloads are starting to split between general-purpose clouds and providers with specialized setups. Last week, we covered Verda, a Helsinki-based company expanding its renewable-powered AI cloud—another example of infrastructure tailored for AI. CoreWeave’s work with Rubin suggests it’s positioning itself for inference, where low latency and high throughput matter more than in training.

The validation also puts pressure on other cloud providers. CoreWeave has worked closely with Nvidia, but bigger players like Google Cloud are investing in their own AI infrastructure. Google’s recent moves in India, pairing Gemini models with its cloud, show how general-purpose providers are trying to keep up. If CoreWeave’s testing leads to faster adoption of new hardware, competitors may need to speed up their own AI cloud plans.

For startups and enterprises, the message is that the AI cloud market is changing. Training workloads might still run on hyperscalers, but inference could shift toward providers with infrastructure built for it. CoreWeave’s Rubin testing is one sign of how the space might evolve—with clouds competing on performance, not just size. Whether it can stay ahead depends on how quickly others respond.

Sources: siliconangle.com