Elon Reveals Plans to Power-Up xAI, Open Source AI Models

Elon reveals exciting new plans for xAI, focusing on an impressive increase in computing and an open-source approach to the AI algorithms.

Elon Musk has unveiled bold plans regarding the xAI initiative, unveiling plans to implement 50 million Nvidia H100 chips within the next five years. This roadmap is a huge increase in xAI’s capacity, which is aimed at positioning it as a key participant in the AI computation race.

Why Nvidia’s H100 Chips?

Nvidia H100 chips have become the industry standard for AI projects and are renowned for their incredible processing power, which is around 900 gigabytes of data per second. It is unrivalled for the training of large models of language. These chips are essential to process trillions of parameters fast and accurately, which is why they are the foundation of modern AI projects.

The H100 remains unmatched in performance for AI tasks, creating demand that far exceeds Nvidia’s supply. Consequently, leading AI projects are racing to stockpile them, despite the significant investment required at over $30,000 per unit.

At present, xAI operates roughly 200,000 H100 chips inside its Colossus Data Center in Memphis and is currently working on plans to build Colossus II to further boost capacity. OpenAI reports that it has an identical setup comprising around 200k H100s. Meta, however, is hoping to beat both by setting an aim of 600k H100s in its forthcoming AI infrastructure, which is an important lead in raw computing power.

Although Meta’s systems are not fully operational, their intention to invest hundreds of billions more into AI technology demonstrates an aggressive, long-term strategy to lead AI development.

The Race Beyond Nvidia

It is worth noting that all the major AI companies, including xAI, OpenAI, Meta, and Google, are developing their own AI chips that will reduce dependence on Nvidia as well as improve efficiency. These private projects remain secret, but they may significantly increase capacities and improve performance beyond the current H100 implementations.

While Musk is confident in xAI’s plan to outdo the competition, industry experts observe that Meta is currently best placed to dominate the market for computing capacity by expanding its projects at speed. However, capacity alone does not ensure AI the top spot; new and innovative solutions, as well as strategic alliances, will be essential to the success of the company.

The Open Source Commitment of xAI

Alongside its compute goals, Elon announced plans to open the source of the entire xAI models. This week, Grok 2.5 was released on Hugging Face, which allows users to access the model’s weights and processes. Musk assures that every Grok version will be released, and aims to improve transparency and increase trust in the xAI’s AI development.

However, there are questions about the authenticity of these publications. Considering that Grok 4 specifically checks in on what Elon thinks about certain issues and weighs in on the responses. 

The extent to which these biases are visible when using the free versions is undetermined, leading to suspicions that the models released could be cleaned versions, rather than fully transparent.

Final Thoughts

Elon Musk’s xAI is rapidly growing, becoming a major part of his X Corp ecosystem. While reaching scale in computing and openness are crucial aspects, the long-term impact of xAI will be determined by its capacity to come up with innovative applications and navigate an extremely fiercely competitive AI landscape.