TAIPEI (Taiwan News) — Nvidia on Monday (March 18) unveiled its newest generation of artificial intelligence (AI) chips at its developer’s conference in San Jose.

Nvidia announced the new Blackwell GPU, which it said provides better AI training performance and energy efficiency compared to its Hopper GPU predecessor, according to Nikkei Asia. “Hopper is fantastic, but we need even bigger GPUs,” said Jensen Huang (黃仁勳), co-founder and CEO of Nvidia, during his keynote speech.

“Generative AI is the defining technology of our time. Blackwell is the engine to power this new industrial revolution,” Huang added. Nvidia said many companies are expected to adopt Blackwell, including Amazon Web Services, Dell Technologies, Google, Meta, Microsoft, OpenAI, Oracle, Tesla, and xAI.

Huang noted that it would only take 2,000 Blackwells and four megawatts of power to train a version of the GPT model that powered ChatGPT compared to the 8,000 Hoppers and 15 megawatts to do similar training for the same amount of time, per Nikkei Asia.

The Blackwell GPU is large and combines two separately manufactured dies into one chip manufactured using Taiwan Semiconductor Manufacturing Co.’s (TSMC) 4 nm process (4NP), according to CNBC.

The Blackwell GPU will also be available as an entire server called the GB200 NVLink 2, which combines 72 Blackwell GPUs and other Nvidia hardware made to train AI models, per CNBC. The first Blackwell chip named “GB200” is expected to ship later this year.