Elon Musk’s artificial intelligence venture, xAI, is undergoing a period of rapid change as it pushes ahead with ambitious long-term plans, including its recent merger with SpaceX and proposals to build AI data centres beyond Earth. At the same time, the company has witnessed a steady stream of senior departures, with half of its original founding team now having left the organisation.

On February 10, Yuhuai “Tony” Wu, who led xAI’s reasoning research, announced his decision to step down in a post on X (formerly Twitter). Wu indicated that he was moving on to pursue a new AI-focused initiative, describing the moment as an opportunity for smaller teams empowered by artificial intelligence to create outsized impact. “It’s time for my next chapter,” he wrote, pointing to what he called an era of wide-ranging possibilities.

Within 24 hours, on February 11, another senior figure followed. Jimmy Ba, who oversaw research, safety and enterprise work at xAI, also confirmed his exit. His farewell message suggested a personal reset, saying it was time to “recalibrate my gradient on the big picture”.

Also Read: xAI raises $20 billion to expand AI infrastructure and reach billions of users 

Founding team reduced to six members

The departures of Wu and Ba mean that xAI now has only six of its original 12 founders still at the company since it was established in 2023. Five of those exits have taken place in less than a year.

Earlier in February 2025, Christian Szegedy, a veteran of Google, left the Musk-led firm. Igor Babuschkin departed in August last year to set up a venture firm, while former Microsoft researcher Greg Yang stepped away last month citing health reasons. The first founding member to leave was infrastructure lead Kyle Kosic, who exited in 2024 to join OpenAI.

Broader exodus among researchers

The churn has not been limited to xAI’s founding leadership. Reports suggest that more than half a dozen researchers have resigned in recent weeks.

Among them is Vahid Kazemi, who had previously praised xAI as being far more capable than the OpenAI team he once worked with.

In a post on X, Kazemi said he was tired of ‘boring’ AI work and had grown disillusioned with the direction of AI research, describing much of the work across labs as repetitive and lacking originality. “All AI labs are building the exact same thing, and it's boring,” he wrote, adding that he was starting something new to explore more creative possibilities.

According to a report by the Financial Times, some current and former employees have pointed to internal friction within xAI. These include concerns that company leadership overstated technical progress to Musk in an effort to close the gap with rivals such as OpenAI and Anthropic, leading to unrealistic expectations. Coding tools developed internally have also struggled to match competing offerings, including Anthropic’s Claude Code.

Also Read: xAI acquires third US site as Musk targets almost 2 GW AI compute 

Public scrutiny and pressure to perform

xAI has also faced criticism over how its AI models are deployed on social platforms, including controversy surrounding Grok and the so-called “undressing” incident on X. Meanwhile, AI companion products such as the anime-style character Ani have failed to attract the level of user engagement Musk had anticipated.

In response, xAI has begun reshaping parts of its organisation. Manuel Kroiss, a former Google DeepMind engineer, has been promoted to help oversee coding operations. Alongside these changes, pressure is mounting for the company to demonstrate financial momentum, with Musk expected to take the merged xAI–SpaceX entity public as early as June.