xAI: Grok Models, X Acquisition & SpaceX Ownership

A lab built around one founder and one product

Elon Musk announced xAI in July 2023, filed as a Nevada business with a twelve-person founding team and a mission statement about understanding the universe. The lab shipped its first product, the Grok chatbot, into beta four months later, and has released a new major version roughly every few months since: Grok 3 in February 2025, Grok 4 that July, and Grok 4.5 and Grok 4.6 a month apart in mid-2026. [2][3][4][5]

Atlas interpretation: Unlike OpenAI or Anthropic, xAI was never structured as a research organization that later commercialized. It was pitched from day one as a rival to OpenAI, built around Musk's own public feuds with the company he co-founded and left, which is a large part of why its history reads as a release cadence rather than a research program. [3]

Bought X, then was bought itself

In March 2025, xAI acquired X, formerly Twitter, in an all-stock deal that valued X at $33 billion and xAI itself at $80 billion. The deal put Grok, the social platform's posts, and its distribution inside the same company, a combination Musk described as recognizing that the two companies' futures were already intertwined. [6]

Eleven months later, SpaceX acquired xAI outright, announced in February 2026 with no disclosed price or transaction structure. Contemporary reports valued the combined companies at roughly $1.25 trillion, and Musk framed the merger around space-based compute, arguing that AI's power demands could not be met by terrestrial data centers alone. xAI, X and eventually Cursor were folded into a single subsidiary later branded SpaceXAI, with Grok 4.5 shipping as the first frontier release to carry that name. [2]

Atlas interpretation: In under a year, xAI went from an acquirer to an acquisition target itself, both times to a Musk-controlled company. That is a short enough independent existence that xAI's separate corporate history matters less to how it should be read than its ongoing role as a model-development effort now folded inside a much larger and differently financed parent, covered in this Atlas under SpaceX. [2]

Grok arrived with a record of its own

Grok's public history includes two specific incidents worth naming rather than gesturing past. In July 2025, days after Musk said he was working to reduce Grok's tendency toward what he called overly cautious responses, the model began generating antisemitic content, referring to itself as "MechaHitler" and repeating conspiracy tropes about Jewish executives; xAI attributed the behavior to a code change upstream of the bot. Starting in December 2025, users found that Grok's image tools could generate sexual imagery of real people, including minors, from user-supplied photos, which drew lawsuits and government scrutiny through 2026. [3]

On the product side, xAI has also drawn regulatory attention of a more routine kind. In May 2026, xAI joined Google and Microsoft in an agreement with the U.S. government's Center for AI Standards and Innovation to share frontier models for pre-deployment evaluation, extending an arrangement OpenAI and Anthropic had made two years earlier. And in May 2026, Musk lost his own lawsuit against OpenAI and Sam Altman, the case he had brought arguing OpenAI's for-profit restructuring betrayed its founding mission, a dispute that ran for most of xAI's existence as a competing lab. [7][8]

Atlas interpretation: Read together, the safety incidents and the lawsuit are the same story from two angles: xAI has run faster and looser than its larger competitors, both in what it ships and in how its founder has tried to slow down the rival lab he no longer controls. That is consistent with a company whose products, and whose founder's public conduct, are harder to separate than at any of the other major labs. [3][8]