Here's what Nvidia announced at GTC 2025
The post Here's what Nvidia announced at GTC 2025 appeared first on Android Headlines.


Nvidia just held its GTC AI conference, and we got a taste of what AI has in store for the future. Here’s what was announced during Nvidia’s GTC conference.
GR00T N1 model
Starting off with the most interesting announcement, Nvidia announced a new foundation model that is specifically designed for humanoid robots called GR00T N1. It’s an open-source model trained on both existing and synthetic data. Synthetic data is data that’s been generated from AI.
This model will use both fast and slow thinking, which mimics how humans think. It will use faster thinking to translate planned action into robotic movements. This is similar to how people can move their arms and legs without needing to think about it. When a robot needs to reason about its environment, then it’ll use slower thinking. It needs to process the visual data in order to know where it is.
New Blackwell Ultra
Last year, Nvidia introduced the Blackwell AI chip, and it was a pretty big deal at the time. During the conference, the company took the wraps off of its latest and greatest AI chip called the Blackwell Ultra. This chip is designed for reasoning models, which the company’s CEO, Jensen Huang, says is the future of AI.
A reasoning model is different from a regular AI model in that it lists the entire thought process as it generates a response. OpenAI’s o1 is an example of a reasoning model along with Gemini 2.0 Thinking and DeepSeek’s R1.
The Blackwell Ultra will fit into Nvidia’s NVL72 rack server. This is the same rack that the first Blackwell chip fit into. The much more powerful chip should bring in some serious money for the company. According to Nvidia, the first Blackwell chip brought in $11 billion in revenue.
New Vera Rubin chips
This year marks a pretty significant milestone for Nvidia. We all know it as the company that makes some of the best GPUs on the market. However, its new Vera Rubin chips see Nvidia introducing its own custom CPU as well.
Vera Rubin is a new system developed by the company slated to launch sometime next year. Vera is the new CPU, and Nvidia says that it’s twice as fast as the CPU used for its Grace system last year. As for Rubin, this is a GPU; rather, it’s two GPUs that work in tandem. This plus the Vera CPU should make this system a force to be reckoned with.
Nvidia says that it can manage 50 teraflops while doing inference. This dwarfs what we saw with the company’s current Blackwell chip, which is 20 teraflops.
The post Here's what Nvidia announced at GTC 2025 appeared first on Android Headlines.