San Francisco, September 9, 2025 – Nvidia has once again raised the bar in artificial intelligence hardware with its announcement of Rubin CPX, a new AI chip engineered specifically for heavy-duty tasks like video creation and software generation.
With expectations for launch by late 2026, this next-generation GPU is built to tackle the challenges of massive context loads—up to 1 million tokens per hour of video—by integrating video decode/encode and AI inference into a single, ultrafast package.
More importantly, Nvidia projects that a $100 million investment in this infrastructure could unlock as much as $5 billion in token-driven revenue.
The Power Behind Rubin CPX: Scaling AI to New Heights
Emerging from the Vera Rubin NVL144 CPX rack-scale system, Rubin CPX offers unprecedented compute density—boasting 8 exaflops, 100 TB of memory, and lightning-fast bandwidth. Nvidia says this delivers a stellar 7.5× performance gain over its previous Blackwell-based systems.
This isn’t just about raw power: Rubin CPX is purpose-built for long-context inference, meaning it can process massive swaths of data—like entire videos or sprawling codebases—with far greater efficiency.
As TechCrunch notes, it supports AI workloads with context windows larger than 1 million tokens, perfect for video generation or AI-assisted coding.
Why It Matters: Real-World Impact
- Studios and content platforms could finally streamline high-quality, long-form video generation—imagine autonomous editing or instant highlight reels made by AI.Developer tools can harness fuller context to generate code that spans entire projects, not just short snippets—ushering in a new era for intelligent coding assistants.Monetization models in AI-as-a-service may tilt toward token-based billing, with Rubin CPX acting as the engine behind scalable, high-revenue APIs.
The Broader AI Landscape
Nvidia’s push comes amid a flurry of strategic industry shifts:
- Its Rubin GPU and Vera CPU, critical components of this platform, are already in the tape-out and fabrication stage at TSMC, pointing to serious momentum toward 2026 deployment.Meanwhile, Nvidia has clarified that despite tight demand, its H100 and H200 GPUs are not sold out, maintaining healthy supply levels.And over in Europe, Germany just activated the Jupiter exascale supercomputer, powered by Nvidia tech—a strategic nod to regional ambition in high-performance AI research.
Why You Should Care
- Game-Changing Context Handling – Rubin CPX’s ability to process enormous context windows could reshape the foundations of generative video and AI coding.Next-Level Efficiency – Embedding video decoding and encoding within the GPU removes bottlenecks, enabling seamless workflows from data input to intelligent output.Ecosystem Acceleration – From creators and filmmakers to enterprise software vendors, access to Rubin CPX could unlock innovations that redefine productivity.
In Summary: Nvidia’s Rubin CPX isn’t just another GPU—it’s a strategic leap into a world where AI handles long, complex tasks with efficiency and scale. As companies prepare to tap into this next-gen infrastructure, the payoff may well be transformative for industries across the board.
