Nvidia unveils Rubin CPX GPU for million-token AI processing

SANTA CLARA - Nvidia (NASDAQ:NVDA), the semiconductor giant with a market capitalization of $4.07 trillion and impressive revenue growth of 71.55% in the last twelve months, announced today its new Rubin CPX GPU, designed specifically for massive-context AI processing capable of handling million-token software coding and generative video applications. According to InvestingPro data, Nvidia maintains excellent financial health with strong profitability metrics, supporting its continued innovation in the AI space.

The Rubin CPX works alongside Nvidia Vera CPUs in the new Vera Rubin NVL144 CPX platform, which delivers 8 exaflops of AI compute power—7.5 times more performance than Nvidia’s GB300 NVL72 systems—with 100TB of fast memory in a single rack. With a P/E ratio of 47.49, investors are pricing in significant growth expectations for Nvidia’s AI innovations. For deeper insights into Nvidia’s valuation and growth prospects, check out the comprehensive Pro Research Report available on InvestingPro.

"Just as RTX revolutionized graphics and physical AI, Rubin CPX is the first CUDA GPU purpose-built for massive-context AI, where models reason across millions of tokens of knowledge at once," said Jensen Huang, founder and CEO of Nvidia.

The new GPU features 128GB of GDDR7 memory and delivers up to 30 petaflops of compute with NVFP4 precision. It also provides three times faster attention capabilities compared to previous systems, enhancing AI models’ ability to process longer context sequences without performance degradation.

According to the company, the Vera Rubin NVL144 CPX platform could enable businesses to generate $5 billion in token revenue for every $100 million invested. This ambitious projection aligns with Nvidia’s strong financial performance, as the company generated $165.22 billion in revenue over the last twelve months, demonstrating its ability to deliver on high-value enterprise solutions.

Several AI companies are exploring applications for the new technology. Cursor CEO Michael Truell said the GPU will deliver "lightning-fast code generation and developer insights," while Runway CEO Cristóbal Valenzuela noted it represents "a major leap in performance" for video generation workflows.

The Rubin CPX will be supported by Nvidia’s complete AI stack, including the Dynamo platform for AI inference and compatibility with Nemotron multimodal models.

Nvidia expects the Rubin CPX to be available at the end of 2026, according to the company’s press release statement. With the company’s robust current ratio of 4.21 and moderate debt levels, Nvidia appears well-positioned to fund the development and launch of this next-generation technology. For exclusive access to more financial metrics and 21 additional ProTips about Nvidia, visit InvestingPro.

In other recent news, Nvidia reported revenue of $46.7 billion, surpassing its original guidance by $1.7 billion and exceeding consensus estimates by $700 million. This achievement comes despite export restrictions on its H20 GPUs in the Chinese market, which had an estimated $4 billion impact during the quarter. Nvidia’s venture capital arm has also invested in quantum computing company Quantinuum, valuing the firm at $10 billion as part of a $600 million funding round.

Analysts have been active in adjusting their outlooks for Nvidia, with Craig-Hallum raising its price target to $245 from $195, citing strong AI demand trends. Benchmark also increased its price target to $220 from $190, maintaining a Buy rating after the company’s quarterly results. Meanwhile, UBS reiterated its Buy rating and $205 price target, despite noting mixed quarterly results and guidance that slightly missed some investor expectations. JPMorgan has reiterated an Overweight rating, highlighting continued strong demand for Nvidia’s data center products. These developments underscore the dynamic environment Nvidia is navigating, with analysts closely monitoring its performance.

This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.