Seonglae ChoA line of servers and workstations produced by NVIDIA specialized in using GPGPU to accelerate deep learning applications
DGX Usages
After AlexNet in 2012, 2014–2016 marked the first deep learning boom as the entire industry shifted to CUDA. Jensen Huang recognized this potential and connected with Professor Hinton's team. After massive investment at NVIDIA, they unveiled the DGX-1 deep learning supercomputer at the GTC 2016 event three years later. However, the market didn't understand that "NVIDIA was becoming an AI company" the audience was quiet, there were no purchase requests, and the stock price plummeted.
Elon Musk approached them at that moment, saying his non-profit wanted to buy one. A non-profit buying a $120,000 supercomputer? Eager to make any sale, they boxed one up at headquarters and rushed it to San Francisco. Peter Abeel, Dario Amodei, Ilya Sutskever – the world's top young geniuses were gathered in a cramped second-floor office, and Jensen opened the box to deliver it. This was the beginning of everything. That photo of delivering the GPU to OpenAI. Elon looking down with his arms crossed, smiling.
Anyone can imagine how grateful Jensen must have been to Elon. In October 2025, Jensen delivered the latest model, DGX-Spark, to SpaceX – a device with compute power similar to DGX-1 but now shrunk to the size of a book.
As Jensen said, xAI will become the company that builds data centers with new chips the fastest. In this respect, xAI has an advantage.
