Starlink veteran Nicolls takes over xAI to chase Claude parity

Elon Musk has quietly handed the operational reins of his chatbot company to a Starlink lifer, and the choice tells you what he thinks xAI's problem actually is. According to Bloomberg, Michael Nicolls, a longtime SpaceX executive who ran Starlink engineering, took over as xAI's president in April 2026, and the bulk of the engineering organization now reports to him. In an opening memo to staff, Nicolls reportedly wrote that the near-term goal is to "match performance of Claude" and to "make Grok maximally useful." Note what is not in that sentence: a research vision, a scaling-law bet, a moonshot benchmark. It is a delivery goal.

That framing matters because Nicolls is not an AI researcher, he is a systems operator, and the reporting makes clear xAI has been slowed less by a shortage of ideas than by a shortage of execution. Bloomberg and follow-up coverage describe an inconsistent strategy, a pre-training team reduced to fewer than five people, and a run of senior departures including the head of post-training and a Mistral cofounder who had joined earlier in the year. Nicolls has reportedly split engineering into five lines (pre-training, post-training, tools, Grok Code, and video and image), the rough shape Anthropic, OpenAI and Google already run. An internal memo also cited xAI's model FLOPs utilization at about 11%, versus a 35% to 45% industry range, with a stated target of 50%. That is not a mission statement, it is a factory-floor number.

The context for all of this is the corporate reset. SpaceX acquired xAI in an all-stock deal in February 2026, valuing the combined entity at roughly $1.25 trillion, and SpaceX went public in June 2026. So the identity-crisis framing in the Bloomberg piece has real weight behind it: xAI has been folded into a newly listed space company whose investors will now be looking at chatbot losses on the same page as launch cadence. Putting a trusted SpaceX operator in charge is the most Musk-shaped response possible to that pressure.

The honest caveat is that most of this rests on internal memos and unnamed insiders, and the turning-a-corner line is a claim from people close to the reset rather than a measurable outcome yet. What the reporting does not give you is any post-IPO segment disclosure on xAI's revenue, spend, or compute cost, nor a clear picture of how much day-to-day product authority Musk himself is actually ceding. If Nicolls does close even part of the gap to Claude while riding SpaceX's balance sheet and compute, the interesting competitor in the coding-agent tier a year from now may not be the one you are watching today.