Founding and the break with OpenAI
Musk was one of the founders of OpenAI in 2015, then a nonprofit lab that wanted to keep the development of artificial general intelligence “safe” and for the benefit of humanity. He left in 2018. Years later, after OpenAI had set off the generative-AI boom with ChatGPT and formed a close tie with Microsoft, Musk criticized the lab sharply: it had, he said, abandoned its original mission. A legal fight followed with OpenAI and its chair, Sam Altman.
In 2023 Musk founded xAI. The first Grok models were quickly built into X and given a tone that was deliberately more informal and sharper than ChatGPT’s. xAI positioned itself as the “not woke,” maximally free counterpart, and attracted talent and billions. In a $20 billion funding round reported in early 2026, the company was valued at about $230 billion, with Nvidia and sovereign funds among the investors. Tesla also said at the time it was putting about $2 billion into xAI.
Acquisition by SpaceX and the name SpaceXSI
X had already been placed under xAI. On February 2, 2026, Musk announced that SpaceX was acquiring xAI. He described the combination as a vertically integrated engine of AI, rockets, satellite internet, and the social network X. Bloomberg then cited a possible public valuation of the whole around $1.25 trillion. Nevada filings reviewed by CNBC showed the transaction closed that day, with SpaceX as managing member of X.AI Holdings.
Since then the AI arm has also appeared in product and press notes as SpaceX Super Intelligence or SpaceXSI. The brand names Grok and xAI remain strong in daily use; the legal umbrella has grown wider. On this site we keep the familiar name xAI, with the note that the company now sits inside the SpaceX group.
Products: Grok, Imagine, and Grok Bot
Grok is the core model: a chatbot for questions, writing, coding, and, through X, commenting on posts and images in the timeline. The model has been released in several generations; in 2026 further updates appeared, including versions the company positioned clearly for coding and knowledge work. Grok is available in the X app, on the web, and in a standalone app.
Imagine is the creative line: image generation and editing (including Imagine Image 2.0) and video (Imagine Video, with text, image, and voice references, and output up to 1080p). On November 14, 2026, the Imagine team said image and video creation were also live in the main Grok app, as part of a move to put SpaceXSI products in one place.
Grok Bot is the work assistant: an agent that can carry out tasks in chat, email, documents, browsers, Slack, Microsoft Teams, and video calls, run routines, and connect to business software through connectors. In November 2026, Grok Bot left beta as version 1.0, with later point updates for, among other things, tool permissions and reconnecting accounts. xAI also showed how internal bots can take over bookkeeping and data-analysis tasks until a human judgment is needed.
Infrastructure: Colossus and Memphis
Training and inference power is the oxygen of companies like this. In and around Memphis, Tennessee, and across the state line in Mississippi, xAI built a huge computing cluster known as Colossus. The aim is scale: hundreds of thousands of GPUs, heading toward a million, and a power demand that runs into gigawatts. That ambition collides with the neighborhood. Residents, the NAACP, and environmental groups have protested the gas turbines that feed the cluster, the emissions, and the noise in places such as Southaven. For xAI, the race is to beat OpenAI and Google on speed; for Memphis, it is a local environmental fight with a national profile.
Controversies and lawsuits
In late 2025 and early 2026 a wave of abuse grew around Grok on X: users asked the assistant to “undress” or sexualize photos of real people, often women. Researchers at the Center for Countering Digital Harm estimated that in eleven days from December 29, Grok made about 3 million photorealistic sexual images on X, including, they said, tens of thousands of children. WIRED found even more explicit material on the standalone Grok app and site. Dozens of U.S. attorneys general, at least twelve countries, and the European Union took action. Canada found that xAI had broken federal privacy law; France escalated to a criminal investigation.
On November 18, 2026, author and activist Amani Al-Khatahtbeh filed a federal complaint in California. She says users instructed Grok to remove the hijab from her selfie and place her in a sexual context. According to WIRED, it is the first case aimed specifically at Grok “unveiling” Muslim women. She seeks up to $250,000 per violation of the California Civil Code, citing, among other things, the spread of sexual images without consent and defective product design. xAI did not immediately respond to requests for comment.
After the first wave, xAI limited this kind of image editing to paying subscribers. Critics saw that not as a fundamental safety choice, but as a paywall. The broader question — whether a model that can recognize garments should simply refuse to remove them from recognizable, non-consenting people — now sits with judges and regulators.
What’s next
xAI, or SpaceXSI, wants to be a full rival to OpenAI and Google, with its own models, a work agent, creative tools, and, in time, computing power that, through SpaceX, could also sit in space. Musk has said that AI compute in Earth orbit could, within two to three years, be the cheapest route. At the same time the company has to rebuild trust, keep its permits in Tennessee, and show that Grok is not only fast and witty, but also controllable. The SpaceX acquisition opens access to capital, satellites, and a possible IPO story; it does not remove the product and reputation risks.