Nvidia (NVDA) arrived at this year’s Computex Taipei tech expo on Monday with a range of announcements, spanning from the creation of humanoid robots to the expansion of its advanced NVLink technology. This technology enables companies to develop semi-custom AI servers based on Nvidia’s infrastructure. These announcements follow Nvidia’s recent momentum, particularly after the US decided to abandon the Biden administration’s AI export restrictions that would have limited which countries could purchase Nvidia’s AI chips. Nvidia was also highlighted during President Trump’s visit to Saudi Arabia, where the company committed to supplying several hundred thousand AI processors over the next five years to Humain, an AI startup owned by Saudi Arabia’s sovereign wealth fund. At the event on Monday, Nvidia unveiled the Nvidia Isaac GR00T-Dreams, which the company claims assists developers in generating vast amounts of training data to teach robots various behaviors and enable them to adjust to different environments. CEO Jensen Huang has stated that physical AI represents the next trillion-dollar industry worldwide. To achieve this, Nvidia is focusing on developing the software needed to train and operate humanoid robots in factories, with the goal of eventually bringing them into homes. Beyond its robotics initiatives, Nvidia introduced its new NVLink Fusion product, allowing clients to construct custom servers using Nvidia’s Grace CPU alongside a third-party AI chip combined with Nvidia’s server infrastructure.

Customers also have the option to pair their own CPUs with Nvidia’s AI chips. “Using NVLink Fusion, hyperscalers can collaborate with the NVIDIA partner ecosystem to integrate NVIDIA rack-scale solutions for seamless data center deployment, ” the company said in a statement. The aim is to provide infrastructure customers with greater flexibility in developing their data center and server architectures. Additionally, Nvidia is developing its RTX Pro Blackwell servers. These servers, powered by the company’s Blackwell Server Edition GPUs, are intended to facilitate “the shift from CPU-based systems to efficient GPU-accelerated infrastructure, ” according to Nvidia. The company explains that these systems are designed to handle “virtually every enterprise workload, ” including activities such as design and simulation software, running agentic AI programs, and more.