Elon Musk's SpaceXAI Has Lost More Than 50 Employees Since the SpaceX–xAI Merger

Since the February merger of SpaceX and xAI, the new entity operating as SpaceXAI appears to have suffered a noticeable talent drain. According to media reports, more than 50 engineers and researchers have left the company since February. For Musk's AI business, the problem is not only the number but who is leaving.

Scale of the exodus

The SpaceX–xAI merger was announced on February 2, 2026, and in early May the company took the new name SpaceXAI. Since then, more than 50 employees have reportedly left the combined organization. They include specialists who worked on key areas: AI coding, world models and Grok's voice features.

This does not look like ordinary turnover after a rebrand. It is a noticeable loss of people in the teams that directly affect how fast new models and product features ship.

Another signal is where these people are going. Competitors have already picked up part of the team. According to available data, at least 11 former xAI employees moved to Meta, and at least seven more joined Thinking Machines Lab, Mira Murati's startup. For the market this is an important marker: when other labs quickly snap up strong researchers, the cause is usually not cyclical hiring but competitors seeing scarce talent that is ready to hire and taking it without a long wait.

A vulnerable core

The situation in the pre-training team is particularly alarming. Pre-training is the first, foundational stage of building new AI models: it sets the architecture, the quality of the model's baseline behavior and the ceiling for later fine-tuning, and it also involves allocating compute among teams. If such a group shrinks to a handful of people, it immediately hits the company's ambition to build frontier models in-house rather than merely catch up with the market on applied features.

According to reports from inside SpaceXAI, after the head of this area, Juntang Zhuang, left, the pre-training group shrank to an almost symbolic size. Coming after the leadership change that already followed the merger, this raises further questions about the company's priorities. From the outside, it looks as if SpaceXAI is trying to sustain an aggressive development pace, restructure management and keep its focus on its own base models all at once. Doing all three without disruption is usually very hard.

Why people are leaving

There seems to be no single explanation for the exodus. It is more likely a combination of factors: pressure after the deal, team reshuffles, active poaching by competitors and changed financial incentives. Inside fast-moving AI companies, such causes rarely act alone. When several of them stack up, even strong teams start losing people faster than they can replace them. That appears to be what happened here.

• extreme work culture and a high risk of burnout
• unrealistic deadlines for model training
• leadership change after the SpaceX–xAI deal
• aggressive poaching by Meta and Thinking Machines Lab
• weaker incentive to hold on to shares amid tender offers and IPO expectations

Employees familiar with the situation link some departures to a harsh management style and inflated deadlines. Under such conditions teams may start cutting corners, which hurts not only morale but the quality of the product itself. This is especially sensitive for Grok: if rushing affects model training, the consequences quickly show up in the quality of answers and the reliability of new features.

There is also a financial factor. SpaceX regularly holds tender offers that let employees sell already vested shares on the secondary market. The market is also waiting for a possible SpaceX IPO. When people get a chance to lock in part of the value of their holdings, retention through stock options works less well. Put simply, if the financial incentive has already been partly realized, putting up with overwork and internal turbulence becomes far less attractive. For the company this is bad timing: options stop holding people exactly when the business needs maximum stability.

What this means

For SpaceXAI this is a test not only of hiring but of its ability to keep critical teams during business integration. If the exodus continues from pre-training and the research core, both Grok's development and the company's attempts to compete with OpenAI, Anthropic and Meta at the base-model level could slow. Given the expectations around SpaceX and Musk's wider ecosystem, the staffing story is quickly turning from an internal problem into a strategic risk.

*Meta is designated an extremist organization and banned in Russia.