Google Loses AI Researchers to OpenAI and Anthropic Ahead of IPO Push

Published by James Harris on June 22, 2026June 22, 2026

Google’s two most prominent AI departures in a week landed on the same news cycle as a capital expenditure forecast that reads more like a sovereign infrastructure budget than a tech company’s earnings guidance. The stock dropped roughly 7 percent on Monday, its steepest single-day fall in about a year, and the sell-off was less about any one event than the collision of three compounding anxieties.

The talent losses are the most visible pressure point. Noam Shazeer, who returned to Google in August 2024 as part of the Character.AI partnership deal alongside researcher Daniel De Freitas, has now left again for OpenAI, less than two years after coming back. The timing is awkward: Google had just used its I/O developer conference to showcase Gemini 3.5 Flash and the Gemini Spark AI agent, and the story that should be about new products is instead about who is walking out. Then John Jumper, a DeepMind VP who helped build AlphaFold and shared a 2024 Nobel Prize with Demis Hassabis, announced he was joining Anthropic after nine years at the company. OpenAI and Anthropic are both moving toward IPOs, which means researchers at Google are now looking at private competitors offering both mission and meaningful upside, a combination that salary alone cannot easily counter. The crypto VC rebound dynamic offers a rough parallel: capital chasing AI deals across 17 or more rounds in a single month signals just how broadly the talent and funding competition has spread beyond any single incumbent.

Google has already restructured once, merging DeepMind and Google Brain in April 2023. Reorganizations tend to signal internal friction before they signal strength.

The spending numbers are the second problem. Alphabet has raised roughly $141 billion in equity and debt financing since October, and projected capital expenditures for 2026 are running between $180 billion and $190 billion, more than double the prior year, with 2027 described as growing further still. Microsoft CEO Satya Nadella’s weekend interview with the Wall Street Journal, in which he argued AI had become commoditized and urged companies to reduce dependence on large AI providers, landed at exactly the wrong moment. Investors already uncertain about the return on AI infrastructure spending heard a major competitor effectively arguing the moat is narrowing. The Accumulation/Distribution Rating sitting at D over the past 13 weeks suggests institutional holders have been quietly reducing exposure well before Monday’s move. The broader concern, flagged increasingly by credit analysts examining technology infrastructure risk, is that firms racing to build at this scale may be locking in costs before the revenue model fully materializes.

Google’s cloud and search ad businesses delivered a strong first quarter, reported April 29, and the IBD Composite Rating of 94 out of 99 reflects that the underlying business is not broken. The question the market is now asking is whether the businesses that are working today can fund the infrastructure bets being made for 2027 without compressing margins to a point that changes the investment thesis entirely.