- Nvidia announces AI chips, including Blackwell Ultra and Vera Rubin.
- Addresses concerns about China’s DeepSeek model.
Nvidia unveiled new chips for artificial intelligence at its annual GTC conference on Tuesday. CEO Jensen Huang announced two key releases: the Blackwell Ultra chip family, which will be available in the second half of this year, and Vera Rubin, a next-generation GPU, expected to launch in 2026.
Nvidia’s business has surged since the launch of OpenAI’s ChatGPT in late 2022, driving a sixfold increase in sales. The company’s GPUs dominate the market for training advanced AI models, making its new releases highly-anticipated among software developers and cloud providers like Microsoft, Google, and Amazon. The companies are watching closely to see if the new chips offer enough performance and efficiency improvements to justify continued investments in Nvidia-based data centres.
“The computational requirement, the scaling law of AI, is more resilient, and in fact, is hyper-accelerated,” Huang said.
The announcements reflect Nvidia’s shift to a yearly release cycle for new chip families. Before the AI boom, the company typically introduced new architectures every other year.
Vera Rubin chip family
Nvidia plans to ship its next-generation GPU system, Vera Rubin, in the second half of 2026. It will feature a custom CPU called Vera and a new GPU design named Rubin, continuing the company’s tradition of naming chips after scientists. Vera marks Nvidia’s first custom CPU design, based on an in-house core named Olympus. Previously, the company used off-the-shelf Arm-based designs.
Nvidia said the custom Vera CPU will deliver twice the speed of the Grace Blackwell CPU introduced last year. Rubin will support up to 288 GB of fast memory and manage 50 petaflops for AI inference – more than double the 20 petaflops handled by Blackwell chips.
Notably, Rubin will consist of two GPUs working together as a single unit. Nvidia plans to follow up with a “Rubin Next” chip in 2027, combining four dies into a single chip which it claims will double Rubin’s performance.
Blackwell Ultra chips
Nvidia also introduced new versions of its Blackwell chips: Blackwell Ultra. The chips are designed to handle more tokens per second, improving the speed at which AI models process data. Cloud providers are anticipated to gain from Blackwell Ultra’s enhanced performance, and Nvidia estimates that the chips could generate up to 50 times more revenue for their owners than the previous Hopper generation, which was released in 2023.
Blackwell Ultra will be available in several configurations, including a version paired with an Nvidia Arm CPU (GB300), a standalone GPU version (B300), and a rack-based version featuring 72 Blackwell chips. According to Nvidia, the top four cloud companies have deployed three times as many Blackwell chips as Hopper chips.
DeepSeek and AI reasoning
Nvidia also addressed the issue of China’s DeepSeek R1 model, which raised questions among investors when it launched in January, as DeepSeek’s algorithms reportedly required fewer chips than comparable US-based models to perform at comparable levels.
Huang described the model as a positive development for Nvidia, noting that DeepSeek’s ability to perform “reasoning” demands more computational power. Nvidia said its Blackwell Ultra chips are designed to handle reasoning models more effectively, improving inference performance for such tasks.
Nvidia’s strategy moving forward
The GTC conference in San Jose, California, attracted about 25,000 attendees and featured hundreds of companies showcasing how they use Nvidia hardware for AI, including General Motors, which announced plans to use Nvidia’s platform for next-generation vehicles.
Nvidia also announced updates to its networking hardware and new AI-focused laptops and desktops, including the DGX Spark and DGX Station, capable of running large models like Llama and DeepSeek.
Nvidia plans to continue naming its chip families after scientists. The chip architecture following Rubin will be named after physicist Richard Feynman and is expected to launch in 2028.
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo. Explore other upcoming enterprise technology events and webinars powered by TechForge here.
