By Michael Kan
Publication Date: 2026-01-06 01:25:00
Sorry PC gamers, but Nvidia didn’t talk at all about graphics cards or gaming at the company’s speech ahead of CES, the annual electronics trade show. Instead, the event was firmly fixated on the company’s newest AI chips meant to help the largest tech companies unleash even more generative AI programs.
For nearly two hours, Nvidia’s CEO Jensen Huang discussed the company’s enterprise business, especially for AI data centers, which makes up nearly 90% of its revenue. The most significant news was that Nvidia has started “full production” of the company’s Rubin platform, the successor to the Blackwell architecture.
(Nvidia)
The company has talked plenty about Rubin before, but Nvidia is now ready to start pumping out the cutting-edge AI chips, which will be packed into servers meant for the newest data centers. The Rubin platform isn’t one AI chip, but six. The main two are the Rubin GPU, which spans 336 billion transistors, along with the Vera CPU, which features 88 “custom Olympus cores.”
(Nvidia)
(Nvidia)
The Rubin GPU itself promises to offer a five times performance increase in AI inference compared to Blackwell, thus lowering the costs and energy demands to run and train AI programs such as chatbots. The company plans on packing hundreds of Rubin GPUs and the related AI chips into servers, effectively selling the systems like supercomputers. All the major AI labs including Anthropic, OpenAI,…
