Nvidia makes a lot of AI chips, but did you know they also have AI engineers who help build AI chips?

During Computex, Nvidia CEO Jensen Huang spoke to the press about the company’s development. When asked how the company plans to go from releasing new chips every two years to every year, he had a surprising answer.

“AI engineers. We make excellent AI chips, which help us train AI engineers, who in turn help us build better chips. By the way, today not only chips but also software compilers are being improved by AI.”

He emphasizes that it’s not just AI, but also a whole team of engineers working on the chips. “While other companies around the world are laying off a lot of people, we continue to look for talent. We are growing very, very quickly.”

According to the Nvidia CEO, the combination of more engineers and the support of AI engineers should enable the switch to an annual rhythm.

Finally, he emphasizes that it is not just about developing AI chips, such as Blackwell, which will be released later this year, and Rubin in 2026.

“We also make CPUs (Grace), Infiniband switches and DPUs (BlueField). All of these chips are needed to reach the entire market. Just making AI chips is far too limited. Software, hardware, everything needed to build an AI supercomputer, the entire ecosystem is important.”

Finally, he emphasizes the most important asset within the Nvidia ecosystem: an architecture. “There is no separate GPU architecture, CPU, network or fabric. Everything is unified and connected by three software standards: CUDA, DOCA and NCCL. In addition, we have built an AI supercomputer. This makes the development of chips more transparent overall.”