Elon Musk's AI lab started 2026 with 11 co-founders, a $250 billion valuation, and the world's largest GPU cluster under one roof. By June 8, every co-founder is gone, 80-plus engineers and researchers have followed them out the door, and the company's HR department is so strained it paused new hiring on June 3. The SpaceX acquisition of xAI, completed February 2, has produced the most consequential talent dislocation in US AI hiring since the 2023 OpenAI board crisis — and the beneficiaries are now legible.
What Happened
The timeline runs fast. SpaceX closed its all-stock acquisition of xAI on February 2, 2026, in a deal that valued xAI at $250 billion and the combined entity at $1.25 trillion. Within days of close, the co-founder departures began.
Tony Wu left on February 10. Jimmy Ba resigned within 24 hours of Wu's announcement. Toby Pohlen, Zihang Dai, and Guodong Zhang followed in late February and early March. Kyle Kosic — a former OpenAI engineer who had been one of xAI's core pre-training operators — departed alongside Greg Yang, formerly of Microsoft Research, and Pohlen, who had spent six years at Google before joining Musk's lab. Manuel Kroiss, who ran xAI's pre-training team and had prior stints at DeepMind and Google, and Ross Nordeen, the lab's de facto technical program lead, were the last to leave, on March 27, 2026. At that point, all 11 original co-founders had exited.
The departures were not quiet. Igor Babuschkin — xAI's former chief engineer who had left earlier — publicly launched Babuschkin Ventures to back AI and agentic systems, effectively signaling that the intellectual core of the founding team was rebuilding elsewhere. Several co-founders cited a specific cultural collision: xAI was founded with a research orientation; SpaceX runs on milestone-based engineering culture. When Musk set what multiple sources described as unrealistic model performance deadlines and the post-merger restructuring dissolved xAI's independent research tracks into SpaceX's operational hierarchy, the researchers left.
By May, The Information reported that more than 50 researchers and engineers had exited SpaceXAI since the merger. Metaintro's tracking, which counts broader staff departures including senior engineers below co-founder level, puts the figure above 80. Both figures predate the June 3 hiring pause Bloomberg reported. SpaceXAI halted new intake for Grok training specialists — roles covering finance, law, STEM, and creative domains — after its HR function could no longer process candidates at the rate they were arriving.
The June pause compounds a pattern that began last year. In 2025, xAI laid off at least 500 generalist AI tutors and data annotators, explicitly to make room for a specialist hiring surge. That surge stalled: Jack Schwaiger, who oversaw medicine, legal, and STEM training at xAI, departed in April 2026; Jeffrey Weichsel, who led finance-focused training, also left.
Why It Matters
The displacement of 80-plus researchers from a single frontier lab in a 120-day window functions as a forced liquidity event for the broader US AI talent market. And the destinations tell you who is winning H1 2026.
Meta's Superintelligence Labs absorbed at least 11 former SpaceXAI employees since February, according to The Information — this on top of the lab's existing researcher acquisition campaign that had already pulled Trapit Bansal, Hongyu Ren, Jack Rae, and Pei Sun from OpenAI and Google DeepMind, respectively. Thinking Machines Lab — the startup founded by former OpenAI CTO Mira Murati — recruited at least seven former SpaceXAI staffers. Anthropic picked up at least two. A smaller cohort started companies.
The market-pricing effect is direct. Meta has been offering packages structured as $100 million to $300 million over four years, with a liquid-cash weighting that traditional RSU structures at Google or Microsoft cannot match. When those offers land on candidates who have just watched their founding team dissolve in 90 days, the acceptance rate climbs. Meta CFO Susan Li said on the Q3 2025 earnings call that employee compensation costs were growing at an "accelerated pace," driven specifically by AI technical hires, and that the full-year recognition of 2025 compensation for those hires would be "the second largest contributor to expense growth" in 2026.
The downstream market signal is measurable. Levels.fyi data through May 2026 shows OpenAI L5 software engineers at $1.15 million total compensation ($336,000 base, $774,000 stock per year) — a number that moved materially after Meta's acquisition campaign made $100 million signing bonuses a documented market event, not a rumor. PwC's 2025 Global AI Jobs Barometer recorded a 56% wage premium for AI skills, up from 25% the prior year — the premium itself doubled in twelve months. AI job postings sit 134% above their February 2020 baseline; total job postings are 6% above that same baseline.
SpaceXAI, meanwhile, is not abandoning operations. The Memphis campus — home to Colossus, the 555,000-GPU cluster that went operational in December 2024 and expanded to 2 gigawatts of power capacity in January 2026 — now employs approximately 3,000 people on site. Musk's post-exodus rebuild strategy is relying on internal SpaceX transfers and loyalists from his other companies. Michael Nicolls now leads the bulk of the engineering team, with hires arriving from Google and the AI coding tool Cursor. The lab retains the compute — 555,000 GPUs, 2 gigawatts. What it does not have is the founding research team that justified the $250 billion valuation.
What's Next
Three things to watch through the end of H1 and into Q3.
Grok's next model release becomes a credibility test. SpaceXAI was targeting a new frontier model built from scratch by Q1 2026. That deadline appears to have slipped. With Kroiss — who ran pre-training — gone, and with the specialist Grok training program now paused, the next Grok release will be read by the market as a direct signal of whether the rebuilt team can execute at the level the co-founders enabled. A delay past Q3 2026 would accelerate the talent-flight narrative.
The SpaceX IPO calculus. SpaceX is tracking toward an IPO that Musk's advisors have framed as a $1 trillion event. xAI's talent stability is part of that prospectus story. If the May TechCrunch and The Information reporting on 50-plus departures gets cited in S-1 risk factors, the IPO timeline could compress or the xAI-as-AI-division narrative could be reframed before roadshow. Bankers will be watching the June and July hire-versus-departure ratios closely.
Where the displaced researchers land next. Eleven went to Meta. Seven to Thinking Machines. Two to Anthropic. That leaves a cohort still unplaced — or at minimum unannounced. The second half of June will likely produce a round of LinkedIn announcements from former SpaceXAI researchers. The destinations will function as a real-time league table of which US frontier lab has the most pull in a market where every senior AI researcher is fielding inbound at multimillion-dollar floors.
The H1 2026 US AI hiring story in one line: one lab's structural collapse is funding three other labs' research buildouts. The SF talent market does not mourn vacancies — it routes around them in weeks, not quarters.