Elon Musk told investors this month that his xAI startup plans to build a new supercomputer by fall 2025. According to The Information, this supercomputer will be used to support a more advanced version of the Grok chatbot. Musk describes this supercomputer as a “gigafactory of computing power.”
Tens of thousands of NVIDIA H100 GPUs will be needed to build this supercomputer and its cost will reach billions of dollars. Musk stated that at least 100,000 chips are needed for the third version of Grok. That’s five times more chips than the 20,000 GPUs used to train Grok 2.0.
The planned GPU cluster will be at least four times larger than any structure used today by xAI competitors, Musk said in his presentation to investors. Grok is currently at version 1.5 and this version was released in April. In addition to text, Grok 1.5 can also process visual information such as photos and diagrams.
This supercomputer will provide a great advance for new versions of Grok. XAI began rolling out AI-based news summaries powered by Grok to premium users earlier this month. These developments can be considered an indicator of how far xAI can go in the field of artificial intelligence.
Musk stated that the construction of the supercomputer and the development of Grok required a large financial investment. This supercomputer will make it possible to train faster and more efficient artificial intelligence models. Additionally, xAI is aimed to provide a competitive advantage over its competitors.
The construction of the new supercomputer will create a major infrastructure for not only xAI’s AI chatbots but also other AI applications. This can be described as part of Musk’s vision to take artificial intelligence technologies even further.
Source link: https://www.teknoblog.com/elon-musk-xai-super-bilgisayar-planlari-ortaya-cikti/