NetApp is acquiring DataPelago, a company specializing in AI data infrastructure. The acquisition brings GPU-accelerated data processing to the storage layer. At the heart of the deal is Nucleus, an engine that processes data where it resides, without copying it to separate compute clusters.
According to NetApp, the biggest obstacle to enterprise AI is not the model or the chip, but the preparation, governance, and activation of data. DataPelago addresses precisely that bottleneck by moving computing power to the data layer, rather than the other way around.
With Nucleus, it offers a universal data-processing engine that utilizes both CPUs and GPUs. The engine takes query plans from existing engines such as Apache Spark, Trino, and Ray, breaks them down into individual operators, and executes each component on the most suitable silicon. Applications do not need to be modified for this.
NetApp claims that this approach reduces infrastructure costs by up to 80 percent and delivers performance up to ten times faster than conventional methods.
Integration with AFX and the AI Data Engine
Last year, NetApp introduced a major AI upgrade with AFX and the AI Data Engine. That engine runs on data compute nodes and vectorizes existing data on-the-fly via Nvidia NIM microservices. Nucleus now extends that GPU acceleration to entire data processing pipelines.
With this, NetApp is aiming for what it calls “zero-copy activation”: analytics, GenAI, and agentic workloads run on data where it resides, without duplicate copies.
“NetApp manages more enterprise data across more environments than anyone in the industry,” says Chief Product Officer Syam Nair. According to him, the next phase of AI will be won by companies that let data work at the source.
Following the acquisition, DataPelago will continue as a wholly owned subsidiary of NetApp.