Nvidia announced major advancements in AI and robotics at Computex 2025 in Taipei on Monday. The company introduced new humanoid robot technology and flexible AI server infrastructure options for customers.
One of the main highlights was the Nvidia Isaac GR00T-Dreams platform. This tool helps developers create large sets of training data to teach robots how to perform tasks and adjust to new environments. Nvidia believes these physical AI robots could be a trillion-dollar industry in the future.
Nvidia CEO Jensen Huang said the company is building software to train humanoid robots that will first work in factories, then later enter homes.
Alongside the robot platform, Nvidia launched NVLink Fusion, a solution that lets customers build semi-custom AI servers. Companies can combine Nvidia’s Grace CPU with their own AI chips or pair their own CPU with Nvidia’s GPUs. This gives businesses more control over how they build their data centers.
The company also revealed its RTX Pro Blackwell servers, powered by Nvidia’s new Blackwell GPUs. These servers are designed to handle almost any enterprise workload, including simulations, design software, and advanced AI tasks.
To support global AI development, Nvidia introduced DGX Cloud Lepton, a cloud-based service for AI training and deployment. It will run through partners like CoreWeave, Foxconn, and SoftBank.
Nvidia’s announcements come after months of market uncertainty. The U.S. recently dropped AI export rules that would have limited Nvidia’s chip sales abroad. The company also secured a deal to supply hundreds of thousands of AI chips to Humain, a Saudi-backed startup.
Despite recent ups and downs in its stock price, Nvidia remains one of the top players in the AI and chip industry.
