@xai is building a new Artificial Intelligence training cluster which combines the compute of 100,000 of Nvidia’s best chip (for now), the H100 which retails for around US$30,000 each.
Elon Musk took to @x to announce that the xAI, X and Nvidia teams all worked together to get the new liquid-cooled AI training cluster operational.
At 4:20am, it’s clear the team were working hard to get this to work and with a hardware investment of as much as $3 Billion, its understandable that they’d want to put it to work as soon as possible.
The AI Training Farm is built in Memphis, Texas, around 6.5hrs from Tesla’s Texas Gigafactory.
xAI will use the new training compute to train a new version of @grok which is available to premium subscribers of @x.
Elon recently gave Lex Fridman who has interviewed Musk multiple times, a tour of the new facility.
xAI also shared a photo of the team with the compute cluster in the background.
Musk claims that the new facility is the most powerful AI training cluster in the world.. which does beg the question.. how does this compare with Tesla’s training cluster that is being used to train the model(s) for Full Self Driving software to power the robotaxis and humanoid robots of the future.