On the AI Infrastructure Summit on Tuesday, Nvidia introduced a brand new GPU referred to as the Rubin CPX, designed for context home windows bigger than 1 million tokens.
A part of the chip large’s forthcoming Rubin collection, the CPX is optimized for processing massive sequences of context and is supposed for use as a part of a broader “disaggregated inference” infrastructure strategy. For customers, the consequence can be higher efficiency on long-context duties like video technology or software program improvement.
Nvidia’s relentless improvement cycle has resulted in monumental income for the corporate, which introduced in $41.1 billion in knowledge heart gross sales in its most up-to-date quarter.
The Rubin CPX is slated to be accessible on the finish of 2026.
