The launch was announced Tuesday at Nvidia’s GTC conference.

NVIDIA (NASDAQ: NVDA) founder and CEO Jensen Huang announced the H100 GPU, an AI chip described as the “new engine of the world’s infrastructure.” This massive 80 billion transistor chip uses TSMC’s 4N process and is three times faster than its previous-generation A100 at FP16, FP32, and FP64 compute, and six times faster at 8-bit floating point math. (NVIDIA)

Built on the new NVIDIA Hopper GPU architecture, also announced at the GTC conference, Huang described the H100 as “the biggest generational leap ever — 9x at-scale training performance over A100 and 30x large-language-model inference throughput.”

Huang said the new GPU will help giant “AI factories” move faster.

“AI data centers process mountains of continuous data to train and refine AI models,” Huang said. “Raw data comes in, is refined, and intelligence goes out — companies are manufacturing intelligence and operating giant AI factories.”

Massive demand for AI applications in speech, conversation, and customer service is fueling the need for faster, more sophisticated chips.

“AI is racing in every direction — new architectures, new learning strategies, larger and more robust models, new science, new applications, new industries — all at the same time,” Huang said.

The new Hopper architecture aims to solve for the massive time taken, sometimes months, to train the highly popular Transformer machine learning models. These models have been used for example in OpenAI’s GPT-3 and medical systems like DeepMind’s AlphaFold.

“Transformer model training can be reduced from weeks to days” Huang said.

H100 is in production, with availability starting in Q3.

Nvidia also said it was building Eos, a new supercomputer that will be the world’s fastest when completed.

To be used internally at the company, Eos will be based on the Hopper architecture and comprise about 4,600 H100 GPUs to offer 18.4 exaflops of “AI performance.”

Related Story: “AI Will Never Be Ethical,” Says An AI Tool In A Debate

Image: NVIDIA founder and CEO Jensen Huang