A Story of Rise and Rise#

Imagine that, together with a few acquaintances, you found a non-profit organization to develop artificial intelligence for the benefit of humanity—openly and transparently—with a clear goal as well: to prevent a handful of giant corporations from being the only group on the planet capable of developing and controlling such a powerful technology, thereby protecting their own interests over the public good.

You donate 50 million dollars, help recruit talent. You believe in the mission.

But time passes, and you see that the organization isn’t taking off. You estimate its chances of success at 0% (not even 1%). So you propose merging it with one of your companies to take control and push it forward. The other founders refuse. So you decide to walk away. After all, they’re going to fail anyway… right?

A year later you find out that not only are they alive—they reorganized to include a for-profit subsidiary, raised a billion dollars from the very kind of company they wanted to “compete” against, abandoned the open philosophy, and are now stronger than ever.

Two years later they launch the most viral product of the decade, reach one hundred million users in months, dominate every headline, claim the glory and the prestige… and you are not part of any of it. You could have been there.

I would be tearing out what little hair I have left. But the protagonist of this story doesn’t have time for regrets. The AI fever was just beginning, and Elon Musk was not going to be left out.

Just as the band of the moment is never denied a stage, one of the world’s most successful entrepreneurs is never denied capital. There are leaked messages in which Elon asks for two billion dollars as casually as someone asking a friend for $20.

Seven months after ChatGPT, xAI, his new AI lab, is publicly announced. The original premise was to build an AI that “understood the true nature of the universe.”
Its first model, Grok, launched in November 2023… and it was bad. Which is perfectly normal: a lab’s first iterations are usually nowhere near the state of the art.

In March 2024 came Grok 1.5, and also a vision-enabled version (it could understand images). Still bad. In August, Grok 2 was released: less bad, but still below the level of the leaders. They also presented an image generator that wasn’t technically impressive either, but could generate photos of celebrities and unfiltered content, making it appealing to a certain audience.

Another curiosity: Grok was integrated into the social network X (formerly Twitter), a platform that was later acquired by xAI.

But while the models were improving slowly, one of the company’s greatest achievements took place: the construction of the Colosus data center, with 100,000 GPUs, built from scratch in just 122 days. A project that, according to initial estimates, should have taken two years. But Elon activated founder mode, leveraged contacts and synergies with Tesla, and in four months it was already operational. Three months later, it was expanded to 200,000 GPUs.

That’s where Grok 3 began training, and it was released in February 2025. It debuted first on LMarena and managed to rank among the best models in the world on several benchmarks.

In July, Grok 4 was released—already competitive with the frontier models from OpenAI and Google, and a leader at launch.

In August, they announced code-oriented versions: Grok 4 Code and Grok 4 Fast, cheaper and surprisingly strong. They compete head-to-head with the previously untouchable Sonnet 4 and, in fact, have already surpassed it in market share on OpenRouter. In November, they announced Grok 4.1, a small update that made it competitive with the leading alternatives from Google and OpenAI.

Currently, xAI is building Colosus II, which aims to become the first gigawatt-scale data center.

xAI is the startup with the fastest iteration speed in the recent history of AI. It went from nothing to a frontier lab in just two years.
Its current valuation: 200 billion dollars.