The ambitious integration of xAI into the SpaceX empire was intended to be the final piece in Elon Musk’s grand valuation puzzle, yet the newly minted SpaceXAI is facing a foundational crisis. Since the merger in early 2025, the organization has hemorrhaged more than 50 key researchers and engineers, including nearly the entire pre-training team responsible for the fundamental capabilities of the Grok models. This exodus has left the laboratory’s core technical engine in a state of near-paralysis, precisely as it seeks to compete with entrenched giants like OpenAI and Google.
The departures are not merely numerical; they are surgical in their impact. Aside from Musk himself, all of xAI’s original co-founders have now exited the company. The pre-training division, which once boasted a robust roster of top-tier talent, has reportedly dwindled to just a handful of individuals by May 2026. These losses are being aggressively capitalized on by rivals, with Meta and the Mira Murati-led Thinking Machines Lab emerging as the primary beneficiaries of this sudden influx of elite talent.
At the heart of this breakdown is a clash between Musk’s legendary 'rocket-style' management and the nuanced demands of AI research. Former employees describe an environment of extreme overwork, where mandatory seven-day workweeks and 'unrealistic deadlines' for model training led to corner-cutting that compromised model performance. While Musk’s 'hardcore' culture has propelled SpaceX to historic launch rates, it appears to be backfiring in a field where long-term intellectual rigor and psychological safety are paramount for breakthroughs.
Financial incentives are also playing a disruptive role as SpaceX approaches what could be the largest IPO in history. With the company’s valuation soaring toward $1.25 trillion, long-term employees are finding themselves with significant equity that can be liquidated during periodic tender offers. For many, the opportunity to secure generational wealth far outweighs the appeal of enduring a high-pressure environment where the strategic direction—shifting from model building to 'reconstructing' the team—feels increasingly volatile.
Perhaps the most telling sign of a strategic pivot is the recent multi-billion dollar deal with Anthropic, which has leased out significant compute power from SpaceXAI’s Colossus data centers. This transition suggests that SpaceXAI may be evolving from a front-line model developer into a global infrastructure provider. By acting as a 'landlord' for GPUs, Musk can generate the consistent revenue streams needed to satisfy Wall Street investors ahead of the IPO, even if his own AI models, such as Grok, continue to lag behind market leaders in enterprise adoption.
As the roadshow for SpaceX’s public debut nears, the company faces a fundamental question of identity. Musk has long pitched AI as the 'brain' of his interconnected ecosystem of satellites and robots, but a brain without a core research team is little more than a collection of expensive silicon. Whether SpaceXAI can successfully 'rebuild' under this new infrastructure-heavy model will determine if it remains a serious contender in the intelligence race or settles into a role as the utility provider for the next generation of computing.