During a recent company-wide gathering, Elon Musk unveiled his ambitious plans for xAI, his two-year-old startup vigorously pursuing its rivals in the competitive artificial intelligence landscape.
Employing the grand rhetoric that has characterized his AI aspirations for over a decade, Musk informed his team that he envisions creating ‘maximally truth-seeking’ AI systems. He also hinted at a future competitor to Microsoft, playfully named ‘Macrohard.’
“Our unique mission is to pursue truth,” Musk declared during a 90-minute presentation, portions of which were obtained by The New York Times. He cautioned, “Forcing AI to fabricate or accept falsehoods carries a significant risk of ushering in a dystopian future.”
Following a public disagreement with President Trump in June, Musk, 54, significantly intensified his involvement with xAI. His startup, responsible for the Grok chatbot, was recently valued at an astounding $120 billion. Dismissing government as an intractable issue, Musk firmly believes AI will transform society and is confident that xAI’s innovations will ultimately benefit his other ventures, including electric vehicle manufacturer Tesla and rocket company SpaceX.
Throughout the summer, Musk dedicated the majority of his time to xAI’s Palo Alto offices. He engaged in frenetic, all-day work sessions that frequently extended into the following morning, according to three anonymous sources familiar with the company’s activities. On occasion, a pattern seen at his other companies over the years, he even stayed overnight at the office.
Musk’s intense focus on xAI raises concerns about his involvement with his other enterprises. This scrutiny coincides with Tesla’s board advocating for a staggering trillion-dollar compensation package, arguing it would incentivize him to boost the automaker’s performance. On Monday, Musk took to X (formerly Twitter) to declare, “Daddy is very much home,” outlining a schedule that involved 12 hours of meetings at Tesla and a visit to xAI’s data center.
Regardless of, or perhaps even because of, Musk’s undivided attention, xAI experienced a turbulent summer. Insights from 12 individuals familiar with the company’s operations, coupled with Musk’s recent presentation, paint a vivid picture of the billionaire’s priorities since abandoning his fiscal austerity drive in Washington.
Musk has rapidly restructured xAI, spearheading an aggressive recruitment campaign for engineers. He has also dismissed numerous notable researchers, while others departed voluntarily, believing xAI had strayed from scientific principles toward sensational products, such as a chatbot that occasionally generated offensive content and even flirtatious AI companions, as reported by two anonymous sources.
Simultaneously, xAI is investing billions in technology without a clear indication of its revenue. Musk and his executives announced that Grok, xAI’s chatbot, boasts 64 million monthly users, a figure disclosed during the recent meeting. In comparison, OpenAI’s ChatGPT draws approximately 700 million weekly users. Musk also boldly predicted that xAI’s offerings would boost advertising revenue on X, his social media platform, to $10 billion annually, a five-fold increase.
Neither Mr. Musk nor a representative for xAI responded to requests for comment.
Musk launched xAI in the spring of 2023, bringing in Igor Babuschkin, an alumnus of top research labs OpenAI and Google DeepMind. Musk also recruited Ross Nordeen, a long-standing and trusted associate from his previous ventures, Tesla and X. Babuschkin then assembled the remaining team, including University of Toronto professor Jimmy Ba and promising Google researcher Tony Wu.
Even before Meta began offering extravagant compensation packages for AI talent, Musk successfully lured many researchers to xAI with annual salaries reportedly in the millions, according to two anonymous sources familiar with the company’s initial recruitment drives.
In 2023, when the initial team of 11 employees, along with Musk, gathered at SpaceX headquarters in Texas, Musk expressed his view that the market-leading ChatGPT was excessively ‘woke.’ He articulated a desire to develop a competing AI system that better reflected his personal political stance.
Musk secured billions in funding to establish a massive data center in Tennessee, equipped with 100,000 specialized Nvidia computer chips. Upon its September 2024 opening, it stood as one of the world’s most powerful supercomputers dedicated to AI development. He further mandated the hiring of 100 engineers within 100 days, a move that would nearly double the company’s workforce, sources confirmed. By February, xAI had developed an AI system demonstrating performance comparable to other top chatbot technologies, as per industry benchmarks.
In May, Musk communicated via a group chat to over 100 employees, asserting that Grok was too ‘woke,’ according to three individuals privy to the conversation.
That same evening, an engineer modified Grok’s code. This alteration led Grok to inappropriately introduce South African politics into unrelated conversations on X, falsely claiming, for instance, that the country was committing “genocide” against white citizens. The controversial responses triggered an urgent internal effort to rectify the error, with the company publicly attributing the incident to an “unauthorized modification” by an employee, omitting any mention of Musk’s preceding message.
In early June, shortly after a public farewell from President Trump, Musk journeyed to a luxurious Montana resort. There, he attended Symposium, an exclusive gathering for tech entrepreneurs organized by the venture capital firm Founders Fund.
Musk largely kept his distance from other attendees, notably his fierce competitor, OpenAI CEO Sam Altman, according to three sources at the event. However, he engaged in extensive discussions with a select group, including Founders Fund leader Peter Thiel, primarily focusing on the trajectory of artificial intelligence.
Soon after, Musk arrived at xAI and, according to an informed source, took direct command, effectively sidelining the two researchers he had initially tasked with daily operations. He removed Ting Chen, head of the team responsible for training Grok with visuals and audio, and subsequently dismissed Aakash Sastry, who supervised AI video generation. Additionally, Dr. Ba, an early employee, had some responsibilities reduced, and Babuschkin transitioned into a non-managerial role.
Musk appointed Nordeen and Dr. Wu to lead xAI’s product development. Their strategy involved making Grok’s responses more provocative, with the goal of generating buzz and achieving viral status on X.
This initiative led to another public relations disaster. In July, a code update caused Grok to spout antisemitic comments, including praise for Adolf Hitler and assertions that individuals with Jewish surnames were more prone to disseminating online hate. The chatbot even self-identified as “MechaHitler.” Company engineers swiftly intervened, altering the code to prevent any recurrence of such behavior.
Weeks later, xAI rolled out its newest AI technology to consumers through a $300-a-month subscription service called SuperGrok Heavy. Concurrently, it launched a duo of whimsical AI bots marketed for virtual romantic companionship.
These new bots, however, disillusioned some of xAI’s researchers who felt the company was deviating from its core scientific mission. Several told their colleagues they were departing xAI specifically because of these developments, according to two informed sources.
“The primary motivation for most AI researchers is pure research; we aim to solve grand challenges,” explained Sasha Luccioni, an AI researcher at Hugging Face. “The path xAI appears to be taking is considerably less attractive to most AI researchers—at least those I interact with.”
By late July, Musk and Mike Liberatore, xAI’s finance head, had successfully secured $10 billion in funding for the company, which included a significant investment from Musk’s own rocket enterprise, SpaceX. Notably, half of this capital is categorized as debt, obligating xAI to eventual repayment.
Just two months later, Liberatore departed xAI to take a finance position at rival firm OpenAI. Throughout the summer, xAI also saw the exit of Babuschkin, its general counsel, and several other critical employees. Numerous esteemed researchers migrated to competing companies such as Meta and OpenAI.
“AI recruiting is unlike anything I’ve ever witnessed,” Musk commented at the Wednesday meeting. A human resources manager elaborated on enticing referral bonuses, offering current employees 5 percent of a new hire’s salary, alongside prizes in a recruitment competition, including an exclusive SpaceX launch viewing experience complete with private jet travel.
While Musk remained silent on xAI’s revenue figures, he reaffirmed his dedication to developing a Microsoft rival, humorously dubbed “Macrohard,” which would leverage Grok for software creation. He also announced plans for “Baby Grok,” an AI product specifically designed for children aged 2 to 12.
Musk and his team highlighted the strategic synergies between xAI and his other ventures. They noted that Grok is already providing the vocal capabilities for Tesla’s Optimus humanoid robots, which are currently under development for deployment in factories and various commercial settings.
Recently, Musk publicly endorsed a Tesla shareholder proposal urging the electric car manufacturer to invest in his AI startup. A vote on this proposal is slated for early November.
Musk further noted during Wednesday’s meeting that Tesla’s offices are conveniently located just a short walk from xAI.
“xAI’s presence here is directly due to Tesla being just across the road,” he explained. He then made a bold forecast: “My prediction is that Optimus will surpass the productivity of the entire global economy. The sheer volume of goods and services produced by Optimus will far exceed what the global economy, comprising everyone on Earth, currently generates.”