xAI founder shared AI infrastructure upgrade roadmap

Artificial Intelligence (AI) company xAI founder, Elon Musk shared a roadmap to upgrade infrastructure and improve the computing system for new and existing products.

Musk’s comment came on a post made by Perplexity AI’s CEO Aravind Srinivas. He asked via poll, which company would build the first 1 million H100 GPU data center.

The poll includes Lllama maker Meta, ChatGPT maker OpenAI, Gemini maker Google, and Grok maker xAI (mentioned as X). The poll drew attention from its chief “Given the technology improvement, it’s not worth sinking 1GW of power into H100s”.

The comment makes sense as investing in such infrastructure and quantity won’t be a better idea if the technology advances next year. We can connect this comment to NVIDIA’s Jensen Huang, who revealed the next AI product lineup at today’s COMPUTEX technology conference in Taipei.

Huang said the Rubin platform will succeed the upcoming Blackwell platform including new GPUs, a new Arm-based CPU – Vera, and advanced networking with NVLink 6, CX9 SuperNIC, and X1600 converged InfiniBand/Ethernet switch. Earlier this year, NVIDIA unveiled the Blackwell platform for cloud customers. It is capable of building and running real-time generative AI on trillion-parameter large language models.

xAI Roadmap

Musk said that xAI will reveal its 100k H100 liquid-cooled training cluster in a few months. H100 is dominating the AI industry’s demand perspective. In that case, xAI could introduce these GPUs as the first major infrastructure upgrade.

However, he also shared the next major upgrade for the xAI by launching a 300k B200s GPU-powered cluster with CX8 networking next summer. B200 is a unified AI platform for developing to deploying pipelines. It supports fifth-generation NVLink and uses 8x Blackwell GPUs with 1,440GB total GPU memory.

This GPU has 3 times the training performance and 15 times the inference performance compared to DGX H100. The system can deliver 72 petaFLOPS training and 144 petaFLOPS inference with 14.3kW max power consumption.

Recently, xAI has raised $6 billion in funding and the company said it will use the investment to upgrade AI infrastructure and improve the product lineup.