The world’s first GW-level computing cluster! Musk announces that Colossus 2 under xAI is now operational, less than one year since construction began!

```

On the afternoon of January 17, 2026, Eastern Standard Time, Musk officially announced that xAI's Colossus 2 supercomputer had officially been put into operation.

It is worth noting that Colossus 2 is the world's first AI training cluster with single-unit computing power reaching the gigawatt (GW) level, and it took less than a year from groundbreaking to completion.

According to public data, a sustained power load of 1GW already exceeds the peak electricity consumption of San Francisco, equivalent to the energy usage of a large power plant or a major industrial manufacturing base. Musk also revealed that the cluster will be further upgraded to 1.5GW in April of this year.

In Epoch AI’s industry comparison chart, xAI’s computing power shows a steep upward curve. While competitors like Anthropic and OpenAI still have similar-scale plans on their 2027 roadmap, xAI has already turned this plan into reality today.

Blitzkrieg of Industrial-scale Computing Power

In the AI field, speed often means everything. The most striking aspect of Colossus 2 is its incredible speed from concept to operation.

Looking back at xAI's development journey, the first-generation Colossus facility in Memphis was originally converted from an industrial building and rapidly became the core of the early infrastructure. The construction of Colossus 2 is a further expansion based on this foundation.

According to an independent report from the end of 2025, xAI purchased additional land near the original facility specifically to support this gigawatt-level expansion of computing power. From breaking ground to full-load operation today, the entire process took only about a year and a half. This "blitzkrieg"-style construction speed is extremely rare in the world of supercomputing centers, where cycles are usually measured in years.

This milestone marks that the training of large AI models has moved beyond simple "software laboratories" and officially entered the era of "industrial-scale compute."

For investors, it must be recognized: Large-scale AI model training is no longer just code running on laptops or ordinary server rooms—it has evolved into a supreme challenge to physical reality. Tens of thousands of dedicated GPUs performing matrix computations require massive power supplies, precision liquid cooling systems, and non-blocking high-speed network interconnects.

The completion of Colossus 2 proves that xAI's engineering capabilities in handling hardware integration, energy delivery, and thermal management—these "hard tech" aspects—have become its core competitive barrier.

Musk: Within 5 years, xAI's AI computing power will surpass the total of all other companies

Unlike OpenAI, which relies on Microsoft Azure, and Anthropic, which relies on Amazon AWS, xAI opted for the hardest route: fully self-built infrastructure.

This vertically integrated strategy gives xAI tremendous strategic initiative. By building its own facilities from scratch around computational loads, xAI does not need to adapt to existing data center architectures; at the same time, the company can fully control resource allocation, data flow, and the pace of hardware upgrades.

On the satellite map of the Colossus 2 campus, the roof of one building is painted with the word "MacroHard," which is both Musk-style humor and a naked demonstration of his ambition to challenge traditional software giants like Microsoft. Musk bluntly stated that since software companies do not produce physical hardware, in principle, AI could fully simulate and replace them.

Notably, last December, SemiAnalysis stated that xAI was racing towards the 2GW goal. At that time, Musk boldly claimed that in five years, the total computing power of xAI would exceed that of all other companies combined.

However, this aggressive expansion strategy comes at a cost. Bringing gigawatt-level, high-density computing clusters online means xAI must handle complex municipal, electricity, and environmental issues like a heavy industry enterprise.

In fact, this "get on the bus first, buy a ticket later" approach has already drawn regulatory attention. In January 2026, the US Environmental Protection Agency (EPA) pointed out that xAI used natural gas turbines for power generation at its Memphis base to meet enormous electricity demands, some turbines were operating without the necessary air quality permits.

This article is from WeChat public account "Hard AI". For more cutting-edge AI news, please go here.

Risk Warning and DisclaimerThe market carries risk—invest cautiously. This article does not constitute individual investment advice, and does not consider the unique investment objectives, financial situation, or needs of individual users. Users should decide whether the opinions, viewpoints, or conclusions herein are suitable for their particular circumstances. You invest at your own risk.

```