First Move  ·  AI Software  · 
Inside the labs — Friday morning, 26 June

Google Faces Continued AI Talent Exodus; Anthropic Valuation Soars Past OpenAI

Several key Google DeepMind researchers recently moved to Anthropic, while OpenAI unveiled its first custom AI chip, intensifying the race among top labs.

ShareWhatsAppXLinkedIn

The story

The AI software landscape saw significant shifts this week, with Google experiencing a notable outflow of talent to rivals. Jonas Adler and Alexander Pritzel, both recognized contributors to Google's Gemini model and AlphaFold research, are reportedly heading to Anthropic.

Additionally, Arthur Conmy, another researcher involved with Gemini 2.5 and AI safety, has also joined Anthropic. These departures follow earlier high-profile exits, raising concerns about Google's ability to retain top AI researchers and maintain momentum on its flagship projects, including Gemini.

The moves are partly attributed to the allure of pre-IPO equity at companies like Anthropic and OpenAI, which are preparing for potential public listings. This talent migration underscores the intense competition for specialized AI expertise as labs vie for leadership in model development and real-world applications.

Who moved

OpenAI

What Changed: OpenAI unveiled Jalapeño, its first custom AI chip, developed in partnership with Broadcom.

Consequence: This move aims to lower the cost and increase the speed of AI inference, expanding OpenAI's full-stack platform from models to hardware.

Meta

What Changed: Meta reportedly hired four additional researchers who were previously at OpenAI.

Consequence: This indicates Meta's continued aggressive strategy in acquiring top-tier AI talent to bolster its research and development efforts.

Hang Ten Systems

What Changed: The company, founded by Vishal Sikka, raised a $32 million seed round led by Mayfield and Aramco Ventures.

Consequence: The funding will support the development of AI-first, non-linearly scalable enterprise software development platforms.

Sail

What Changed: Sail secured $80 million in seed and Series A funding led by Kleiner Perkins, valuing the company at $450 million.

Consequence: This capital will be used to optimize AI inference on existing chips, offering a cost-effective solution for running AI faster.

New models

Mythos

Lab: Anthropic

What: An advanced AI model that identified vulnerabilities in classified U.S. government computer systems during controlled testing.

Use: The model's capability was demonstrated in Project Glasswing, an initiative to strengthen critical software infrastructure and evaluate AI risks to national security.

M2.5

Lab: MiniMax

What: A new model released with a benchmark profile positioning it as a cost-competitive alternative to frontier models.

Use: It offers performance suitable for most production tasks at a fraction of the cost of models from other major labs.

Market signals

Anthropic's valuation reached $965 billion, surpassing OpenAI's $852 billion, according to the Hurun Global Unicorn Index 2026.

Implication: This signals an unprecedented concentration of private AI capital in frontier labs and highlights Anthropic's rapid ascent in market perception.

Patronus AI announced a $50 million Series B funding round, led by Greenfield Partners, and unveiled its Digital World Models.

Implication: The funding will accelerate the development of large-scale simulation environments for training and evaluating complex AI agents, reflecting growing demand for AI reliability infrastructure.

Oracle disclosed in an SEC filing that it cut 21,000 jobs explicitly due to AI, while capital expenditures for AI infrastructure surged by 162% to $55.7 billion.

Implication: This indicates a significant shift in corporate resource allocation, with companies simultaneously reducing human headcount and massively investing in AI automation.

What we'll be watching

Reporting + analyst voices: grounded via Google Search at publish time.