The competition among American technology giants is rapidly escalating

Elon Musk stated in his investor speech in May that his artificial intelligence startup xAI is planning to build a supercomputer. Its plan is to use tens of thousands of dedicated computing power chips to form a huge supercomputer, known as the "computing power super factory". Musk expressed his hope to put this supercomputer into operation before the fall of 2025 and personally take responsibility for timely delivery. The new machine will provide power for its next version of the artificial intelligence chatbot Grok.

As one of the co founders of OpenAI, Musk has stated that xAI will compete with market leaders such as OpenAI and Google. The xAI supercomputer currently being developed to support the training of the next generation Grok model. Musk hopes to put this supercomputer into operation before the fall of 2025, and mentioned that training the Grok 2 model requires approximately 20000 Nvidia H100 GPUs, while models above Grok 3 will require 100000 H100 chips. The specific technology of xAI company mainly focuses on developing a chatbot called Grok. Grok is a chatbot that can access social media platform X in real-time, aiming to provide a more intelligent conversation experience through large-scale data processing and analysis.

The founding purpose of xAI is to "understand the true essence of the universe" and plans to work closely with companies such as X Corp and Tesla to achieve its mission. The company has released multiple large models in the past year, including Grok-1, Grok-1.5, and Grok-1.5V, which possess long context comprehension and advanced reasoning abilities, as well as image comprehension abilities. On March 17th of this year, xAI announced the open source of its large model Grok-1, following the Apache 2.0 protocol to open model weights and architecture.

In terms of technological advantages, Musk plans to use tens of thousands of specialized chips to form this giant supercomputer, which are connected by the Nvidia H100. This scale is at least four times that of the current largest GPU cluster. This significant increase in computing power will greatly enhance computing power, enabling it to handle more complex tasks such as long context understanding and advanced reasoning abilities. For example, this type of supercomputer can support the needs of the next generation of AI chatbots, capable of handling long contextual windows of 127K tokens.

According to reports, Microsoft and OpenAI are planning a massive supercomputer construction project, with a total investment expected to exceed $115 billion. This number is more than three times Microsoft's capital expenditure on purchasing servers and other devices last year. Taking Nvidia's newly launched B200 chip as an example, its selling price may be as high as 30000 to 40000 US dollars, which means that even a standard data center would require billions of dollars just to purchase the chip. At this price, approximately 2.5 million B200 chips can be purchased for $100 billion.

In addition, Microsoft and OpenAI's computing power development plan is divided into four stages. The fourth stage is expected to be put into use in 2026, and the fifth stage of supercomputing has been named "Stargate". Microsoft hopes to launch this supercomputing in 2028. The Stargate will be the highest specification supercomputer that Microsoft plans to build in the next six years.

Musk hopes to create a competitive alternative that can surpass Google DeepMind, OpenAI, or Microsoft. This means that xAI's goal is not only to become an industry leader, but also to surpass existing giants in certain aspects.

With the improvement of computer computing power and algorithm optimization, the intelligence level of artificial intelligence will continue to improve. In the future, AI will not only be able to handle more tasks, but also better understand and meet human needs. In the next 5 to 10 years, AI scenarios will be widely used and widely applied. AI will enter almost every field of our lives, including daily life and various professional fields. The value that AI brings to us will be immeasurable!

In fact, we also need to consider the impact of AI on the traditional industry job market. The more work AI does, the less work humans do. Do many basic field tasks require people to do? If AI can complete these tasks at a lower cost, more efficiently, and with higher quality, it will replace many people in positions, leading to an increase in unemployment and more intense competition. AI is a powerful force, but it can be both good and evil.