Whether a company is worth watching depends not only on its current product, but on whether the people behind it are still there. xAI was recently asked that very question: half of its founding team has left.

This piece looks neutrally at the talent drain: exactly who left, how common this is among startups, and whether it really poses a risk to Grok’s next-generation model in training. To get to know SpaceX and its AI map first, see What kind of company is SpaceX.

First, the facts: who left

Let’s lay out the reported facts first.

Per TechCrunch’s February 10, 2026 report, six of xAI’s twelve founding team members had left, exactly half as of that point. The list includes:

  • Kyle Kosic (a former OpenAI engineer; reportedly returned to OpenAI after leaving xAI)
  • Christian Szegedy (a former Google researcher, a key figure behind the Inception architecture and Batch Normalization)
  • Igor Babuschkin (formerly at Google DeepMind and OpenAI; worked on AlphaStar)
  • Greg Yang (formerly at Microsoft Research, a lead on the Tensor Programs theoretical work)
  • Yuhuai Wu (formerly at Google, a researcher on the Minerva math-reasoning model)
  • Jimmy Ba (a University of Toronto professor, co-creator of the Adam optimizer)

Several of them carry weight in both academia and industry, especially Greg Yang, a core figure in theory research. From the list alone, this already reaches the research backbone, rather than the normal turnover of ordinary engineers.

Put it in context: startup norm or warning sign?

The number looks striking, but to judge how serious it is, you have to put it back in startup context.

Founder turnover is fairly common in fast-scaling, high-valuation startups. Some leave after finishing a stage of work, some leave with their reputation and resources to start their own companies, and some leave over differences in direction. Half departing is a share worth noting, but it can’t be equated directly with “the team is empty” or “the company is finished.”

To put it in balance, two readings coexist here: at worst, the loss of the research backbone could create a gap in core capability; at a neutral reading, this is the normal metabolism of a fast-growing company, and what matters is whether the people brought in and those who stayed can carry the baton. Unfortunately the latter is currently a data gap; xAI has not published a full AI leadership org chart, so outsiders can hardly confirm the succession.

The real watch point: Grok 5

Rather than arguing whether “half is serious” or “no big deal,” it’s more useful to fix on a more concrete indicator: the next-generation model.

xAI said as of early 2026 that Grok 5 is still in training. The most lethal place a talent drain can hurt an AI company is its “ability to make the next-generation model.” So whether this drain has real impact is most directly tested by Grok 5’s actual performance after release: whether capability advances on schedule and whether the cadence has been disrupted.

Before Grok 5 actually ships, any claim that “talent left so the model will surely weaken” or “swapping people doesn’t matter” is premature. Treating Grok 5 as the answer sheet to this question is more meaningful than chasing single departure headlines. For how Grok’s model line evolved, see Why Grok went open, then closed.

In closing

xAI losing half its founding team is a signal worth noting, but one that shouldn’t be over-read. It could mean a gap in the research backbone, or it could just be the normal metabolism of a fast-growing company; both possibilities coexist.

What really tells them apart lies in whether core research capability has a gap, not in “how many people left.” And the most direct indicator for testing that is Grok 5, in training. To put this personnel change back in the context of the whole rocket empire, see What kind of company is SpaceX.