SANTA CLARA, 9 September 2025 — Nvidia today unveiled its Rubin CPX chip system, designed to cater to the steep demands of generative AI workloads—particularly video creation and software generation. The launch underscores Nvidia’s dominance in AI infrastructure and propels the company further into high-performance computing leadership.
Built for the multi-modal complexity of modern AI, Rubin CPX unites video decoding, encoding, and inference tasks under one hardware umbrella. This integrated architecture enables efficient handling of workloads requiring up to a million tokens to represent an hour of video, something traditional GPUs cannot manage effectively.
What Rubin CPX Brings to the Table
- Unified chip architecture: By combining video processing and AI inference, Rubin CPX reduces latency and enhances throughput for video and code generation.
- Token-driven economics: Nvidia projects that a US$100 million deployment of CPX systems could yield up to US$5 billion in revenue, pivoting computation into a potent profit engine.
- Strategic timing: Set to launch by late 2026, Rubin CPX complements Nvidia’s existing Blackwell architecture, reinforcing its ability to meet the explosive demand for large-context AI workloads.
Strategic Momentum in AI Hardware
The Rubin CPX heralds a shift from generic GPU pipelines toward specialized processing engines tailored for creative AI applications. As AI continues to pervade areas like video editing, autonomous content generation, and software development, demand for chips like CPX is expected to skyrocket.
Nvidia’s vertically integrated model—spanning from advanced hardware to specialized software—further fortifies this advantage, making it a cornerstone in institutional and enterprise AI strategy. Analysts widely regard CPX as the next frontier in AI hardware monetization.
