Nvidia plans to launch a new AI chip product, Rubin CPX, by the end of 2026.

Rubin CPX will be available as cards for integration into existing servers or as standalone units for data centers.

The company said the chip is intended to improve efficiency in tasks like video generation and software creation by separating the process of understanding input from generating responses.

Nvidia claims that US$100 million worth of the new hardware could enable customers to generate US$5 billion in revenue, though this figure has not been independently verified.

CEO Jensen Huang described CPX as the first chip built specifically for AI models that do reasoning with large amounts of knowledge—millions of so-called tokens—at once.

The new chip is expected to advance software generation capabilities, allowing systems to understand large-scale software projects.

Rubin CPX will also handle video processing tasks, including decoding, encoding, and searching, on a single chip.

.source-ref{font-size:0.85em;color:#666;display:block;margin-top:1em;}

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Task-specific chip design reflects industry-wide shift toward optimization

Nvidia’s CPX design follows a pattern established years earlier when competitors demonstrated significant performance gains through specialization. Google’s Tensor Processing Units, introduced in 2017, ran significantly faster than Nvidia’s GPUs while consuming less energy for specific AI tasks1. The CPX’s approach of separating understanding from response generation mirrors this specialization strategy, suggesting Nvidia has adapted its architecture philosophy to match proven competitive advantages. By creating chips optimized for specific AI workloads like video generation and software creation, Nvidia is responding to the same market pressures that drove competitors to develop custom solutions nearly a decade ago.

Massive AI hardware market growth justifies ambitious revenue projections

Nvidia’s claim that $100 million in CPX hardware will generate $5 billion in customer revenue reflects the explosive growth in AI infrastructure spending. The AI hardware market expanded from $17 billion in 2022 to $125 billion in 2024, demonstrating the scale of investment driving these revenue projections2. Private AI investment in the US reached $109.1 billion in 2024, providing the capital base that makes Nvidia’s ambitious revenue multiples achievable3. Hyperscalers like AWS, Google, and Microsoft have become the largest GPU buyers, creating the concentrated demand that supports high-value hardware deployments2.

Recent Nvidia developments

……