At least 22 professors and researchers have left or taken leave from elite universities this year to join OpenAI, Anthropic, Meta, and Google DeepMind, according to the article. The departures come from institutions including Stanford, Berkeley, and Harvard. The article notes that the transformer architecture that powers modern large language models came out of a Google research paper in 2017, and cites Andrej Karpathy’s move to Anthropic in 2026 as a visible example of this migration pattern. Reasons for leaving include a "compute gap" where training frontier models requires GPU clusters costing hundreds of millions of dollars, which no university grant can compete with, and a compensation gap where senior AI researchers at major labs command packages that dwarf tenured professor salaries by multiples. The article states that this talent migration reinforces an oligopoly dynamic in the frontier model market, could translate into tangible improvements in AI-driven revenue streams for publicly traded Meta and Google, and introduces a long-term risk to the AI ecosystem by hollowing out university research programs.