xAI Restructures amid Exits as Elon Musk Unveils Grand AI and Space Ambitions

In a marathon internal presentation, Elon Musk and senior leaders at xAI addressed the recent wave of departures. Musk described them as layoffs stemming from a shift in organisational structure.

“As a company grows, especially as quickly as xAI, the structure must evolve,” Musk wrote on X. “This unfortunately required parting ways with some people. We wish them well in future endeavours.”

While restructuring is common in fast-growing companies, the scale of the exits—including a substantial portion of the founding team—has raised eyebrows.

Co-founder Yuhuai (Tony) Wu recently announced his exit, writing on X that it was “time for my next chapter”. He was followed by fellow co-founder Jimmy Ba, with Musk thanking him for “bringing us together on this incredible journey” and saying he would remain close to the team as a friend.

Notably, six of the company’s 12 founding members have now left—five within the past year alone. Infrastructure lead Kyle Kosic departed for OpenAI in mid-2024, followed by Google veteran Christian Szegedy in February 2025. Igor Babuschkin left in August last year to launch a venture firm, while Microsoft alumnus Greg Yang exited last month, citing health reasons.

Musk also unveiled a new organisational structure dividing xAI into four core teams: one focused on the Grok chatbot (including voice), another on coding systems, a third dedicated to the Imagine video generator, and a fourth leading the ‘Macrohard’ project—an initiative ranging from basic computer-use simulation to modelling entire corporations.

“[Macrohard] is able to do anything on a computer that a computer is able to do,” said xAI Founding Member Toby Pohlen, who will lead the project. “There should be rocket engines fully designed by AI.”

Deep Dive Into xAI’s Strategy

xAI executives outlined the company’s progress over the past two and a half years, claiming category-leading advances in voice, image and video generation, coding models, and large-scale AI infrastructure, while setting out ambitions that stretch from “Memphis to the Moon”.

The 45-minute video offered significant new insights into Musk’s roadmap for the AI lab, including product strategy, organisational restructuring, and deeper ties to the X platform.

Musk described xAI as “only two and a half years old, basically a toddler,” yet competing with rivals five to 20 years its senior and significantly larger in scale. Despite that, he argued, the company now ranks in the top tier across several benchmarks, including forecasting accuracy, and is generating more images and videos than competitors combined.

Grok remains central to xAI’s strategy. Executives said its voice capabilities were built from scratch in six months and are now integrated into more than two million Teslas. The broader ambition is to evolve Grok into an “everything app” capable of handling legal queries, slide creation, engineering design, and complex reasoning tasks.

Coding has become another strategic frontier. Leaders claimed recent advances have significantly improved the reliability of AI-generated code, enabling internal teams to use Grok Code to develop and refine successive model generations. Musk suggested that AI could ultimately bypass traditional programming entirely, producing optimised binaries directly.

Executives further claimed that xAI’s Imagine tool is generating 50 million videos per day and more than six billion images over the past 30 days, according to internal metrics.

However, those figures are difficult to disentangle from the surge of deepfake pornography that recently proliferated on X. Engagement on the platform rose sharply as AI-generated explicit imagery spread, with an estimated 1.8 million sexualised images produced over nine days alone, suggesting that a portion of the image-generation totals may include controversial content.

In what was the most striking segment of the presentation, Musk reiterated his commitment to space-based data centres despite significant engineering challenges. He went further still, outlining a vision for a lunar factory producing AI satellites, complete with a mass driver—an electromagnetic launch system—to propel them into orbit.

Such infrastructure, Musk argued, could eventually enable AI clusters capable of harnessing substantial portions of the Sun’s energy output, or even expanding beyond our solar system.

“It’s difficult to imagine what an intelligence of that scale would think about,” Musk said. “But it’s going to be incredibly exciting to see it happen.”

Underpinning these ambitions is a vast expansion of compute capacity. xAI’s Memphis supercomputing cluster already houses hundreds of thousands of GPUs, with further build-out underway. Musk maintained that compute scale remains the decisive advantage in AI development, citing NVIDIA chief Jensen Huang’s view that xAI is among the fastest at deploying large-scale AI infrastructure.