Musk’s Colossus 2: 1st AI cluster to exceed 1 gigawatt

Elon Musk has just reached a historic milestone with xAI: Colossus 2 has become the world’s first AI training cluster to exceed 1 gigawatt of power. As a fan of the Musk ecosystem, encompassing Tesla and his parallel projects, this announcement literally blew me away.

We’re talking about an infrastructure that consumes as much as an entire city, a fundraising of 20 billion dollars, and a deployment speed that leaves the competition behind. Impressive on paper, but this technical feat also raises legitimate questions that I will share with you.

Colossus 2: A World Record for AI Power

The announcement has officially been made: Colossus 2 is now operational with 1 gigawatt of power. To give you a concrete idea of what that represents, it’s the equivalent of San Francisco’s peak demand in its entirety.

But this is just the beginning. The planned schedule is even more ambitious:

• Today: 1 GW in operation
• April 2026: upgrade to 1.5 GW
• Final goal: around 2 GW eventually

By combining Colossus 1 and Colossus 2, xAI now has over 1 million H100 GPU equivalents. No other AI player – neither OpenAI, nor Google, nor Meta – has reached this scale of raw power. It’s technically stunning.

An Execution Speed That Outpaces the Competition

What fascinates me even more than the raw numbers is the speed of deployment. For Colossus 1, xAI achieved the feat of going from site preparation to full operation in just 122 days.

When you consider that competitors generally take years to build similar infrastructures, you better understand the competitive advantage. I systematically find this Musk-characteristic execution speed in all his projects.

On X (formerly Twitter), the tech community was quick to react, many highlighting this unique ability to “move fast and build things” on an industrial scale.

The “Tesla” Approach Applied to AI

The parallel with Tesla is striking. This speed directly reminds me of the deployment speed of Gigafactories that we’ve seen in Shanghai, Berlin, or Austin.

The method is identical: rapid iteration, aggressive production rollout, continuous real-time improvement. What works in electric vehicles is applied as is to artificial intelligence. For us, enthusiasts of the Tesla universe, it’s consistent and reassuring.

20 Billion Dollars to Accelerate Infrastructure

To finance this massive expansion, xAI has just completed a Series E funding round of 20 billion dollars, far exceeding the initial target of 15 billion. The list of strategic investors is prestigious:

• Valor Equity Partners
• Stepstone Group
• Fidelity
• Qatar Investment Authority
• MGX
• Baron Capital
• NVIDIA (symbolic: the GPU supplier becomes an investor)
• Cisco

These funds will be allocated to infrastructure development (data centers), products, and AI research. For me, this trust from the world’s largest investors demonstrates that xAI is being taken seriously against established giants like OpenAI or Google.

Grok 5 in Preparation: The Generative AI Race Intensifies

All this computing power is concretely used to train Grok 4 and prepare Grok 5. Recently, xAI has launched several promising products:

• Grok 4 series
• Grok Voice
• Grok Imagine

Compared to ChatGPT (OpenAI) and Gemini (Google), Grok positions itself differently with a unique, less “politically correct” personality. Personally, I appreciate this more direct approach, but the real question remains: will Grok 5 technically live up to expectations?

Grok, Musk’s “Unfiltered” AI

Grok’s main differentiation is its more direct tone and less systematic censorship. Its native integration with X (formerly Twitter) also gives it a unique advantage in accessing real-time data.

For the Tesla audience, accustomed to the disruption and freedom of expression dear to Musk, this approach resonates particularly well.

The Energy and Environmental Stakes of the Project

Now, let’s address the elephant in the room: the energy consumption of 1 to 2 gigawatts eventually. That’s absolutely massive. The legitimate question I’m asking myself, and that many share: what is the energy source used?

The site is located in Memphis, and local controversies have already emerged regarding the environmental impact. As a Tesla enthusiast and strong advocate for sustainable energy, I would really like official clarifications on this point.

There’s an apparent paradox: Musk is the champion of clean electric with Tesla, solar panels with SolarCity, but what about xAI’s environmental footprint? According to industry experts, the energy consumption of data centers represents a major challenge for the ecological transition.

I personally call for more transparency on the energy mix used. If xAI wants to be consistent with the values championed by Tesla, the supply of renewable energy should be an absolute priority.

The technical feat is undeniable: Colossus 2 pushes the boundaries of what seemed possible in AI. It now remains to be seen if this raw power will translate into sustainable technological leadership against OpenAI and Google. And you, what do you think? Does Grok really stand a chance against ChatGPT? Don’t hesitate to share your opinion on X!