Copyright SiliconANGLE News
Data infrastructure giant Vast Data Inc. said today it’s expanding its collaboration with the cloud data center operator CoreWeave Inc., signing a $1.17 billion commercial agreement that fortifies VAST AI OS as the primary data foundation for cloud-based artificial intelligence workloads. CoreWeave has emerged as one of the darlings of Wall Street since going public earlier this year, growing its business exponentially on the back of the AI industry boom. The company operates a vast network of cloud-based data centers that are packed with Nvidia Corp.’s most advanced graphics processing units and other high-end AI accelerators. It rents this computing power to enterprises customers that need more oomph for their AI systems, serving as a direct competitor to public cloud giants such as Amazon Web Services Inc. and Google Cloud. Vast Data is all about data storage, and it has been busy trying to position itself as the primary platform for AI data, which is essential fuel for large language models and generative AI systems. With Vast AI OS, it’s leveraging its “Disaggregated and Shared-Everything” architecture, which separates storage media from compute resources, to create a kind of operating system for AI data. It allows customers to store nearly unlimited amounts of data that can be accessed independently of customers’ computing resources, making it available to AI systems with much lower latency. Vast AI OS supports every kind of storage, including file, object, block, table and streaming data, and offers AI-specific capabilities such as serverless functions and vector search. The platform has emerged as a critical component of CoreWeave’s data center infrastructure, providing the cloud company’s customers with real-time access to enormous datasets, enabling more economic and lower-latency AI training and inference. CoreWeave co-founder and Chief Strategy Officer Brian Venturo said the Vast AI OS underpins many aspects of how the company designs and delivers its cloud infrastructure. “This partnership enables us to deliver AI infrastructure that is the most performant, scalable and cost-efficient in the market,” he promised. Due to Vast AI OS’s scalable system architecture, CoreWeave says it can spin up the platform for customers in any of its data centers, providing them with an immediate, reliable and performant data foundation for the most demanding AI workloads. The expanded partnership will see CoreWeave work with Vast Data to provide even more sophisticated data services to their joint customers, including new features that can help to optimize data pipelines. With these new services, the companies say customers will be able to move even faster and scale more efficiently than before. AI workloads deployed at the kind of scale seen with cloud providers like CoreWeave require a very different approach to storage, said NAND Research analyst Steve McDowell. He told SiliconANGLE that at such scales it’s critical to ensure that GPUs aren’t constantly stopping and starting while waiting on the required data, and so traditional storage systems simply don’t cut the mustard. “This is where specialist providers like Vast Data thrive,” McDowell said. “It has blended the best capabilities of high-performance scalable parallel file systems with AI-optimized features designed to keep AI workflows from waiting on data.” McDowell did say it’s important to note that CoreWeave isn’t wholly reliant on Vast Data for its data infrastructure. It also works with a company called WekaIO Inc., which is Vast Data’s primary competitor, offering customers a choice of either platform. “CoreWeave’s multivendor approach shows that there’s still no single solution to meet the data challenges of AI at scale,” the analyst said. Vast Data founder and Chief Executive Renen Hallak begged to differ, saying the expanded partnership means Vast AI OS will serve as the foundation for the world’s most ambitious AI projects. “Our deep integration with CoreWeave is the result of a long-term commitment to working side by side at both the business and technical level,” he said. “By aligning our roadmaps, we are delivering an AI platform that organizations cannot find anywhere else in the market.” Image: SiliconANGLE/Dreamina AI
