Capital & Industrial Strategy

Anthropic has purchased stealth biotech AI startup Coefficient Bio in a $400 million stock deal, according to The Information and Eric Newcomer. The acquisition represents a strategic pivot from selling general-purpose foundation models to owning the full stack in a specific vertical. Coefficient Bio has been operating in stealth mode, with limited public information about its specific focus area within biotech AI applications. The all-stock structure suggests Anthropic is using its elevated private valuation to make acquisitions rather than deploying cash reserves.

The move follows OpenAI's recent pattern of acquisitions in specific application areas, including its $6.4 billion purchase of Jony Ive's hardware startup and its acquisition of media company TBPN, though CNBC characterizes OpenAI's M&A strategy as 'chasing vibes' rather than following a coherent strategic framework. Anthropic is also reportedly developing its own version of OpenClaw, the agent platform OpenAI acquired, according to Semafor.

Glen Anderson, president of Rainmaker Securities, reports that the secondary market for private shares has never been more active, with Anthropic emerging as the hottest trade while OpenAI loses ground, according to TechCrunch. The heightened secondary activity typically signals investor anticipation of a liquidity event. Both Anthropic and OpenAI are reportedly exploring public listings, according to Axios, which would represent a fundamental shift away from the mega-round private financing model that has characterized AI lab funding.

Anderson notes that SpaceX's looming IPO could reshape the landscape for everyone in the private markets, potentially absorbing liquidity that would otherwise flow to AI companies. The timing is notable given that both Anthropic and OpenAI have raised multiple multi-billion dollar rounds in the past 18 months. A public listing would subject their economics to quarterly scrutiny and require disclosure of revenue, burn rate, and unit economics that have remained opaque in private markets.

Meta, Microsoft, and Google are investing in dedicated natural gas power plants to run their AI data centers, accepting long-term energy price exposure and regulatory risk to secure compute capacity, according to TechCrunch. The decision represents a strategic shift from purchasing power through utility agreements to owning generation assets directly. Natural gas plants can be built faster than nuclear facilities and provide more reliable baseload power than renewable sources, but lock the companies into fossil fuel infrastructure as climate regulations tighten globally.

The capital commitment suggests these companies view energy availability as a greater constraint on AI expansion than compute hardware shortages. Building captive generation also sidesteps grid interconnection queues that have delayed data center expansions by 18-24 months in some regions. However, the plants may face stranded asset risk if future carbon pricing or emissions regulations make natural gas economically unviable before the assets are fully depreciated.

Meta has paused work with Mercor following a security incident that potentially exposed key data about how major AI labs train models, according to Wired. Multiple AI labs are investigating the scope of the breach. Separately, The Wall Street Journal reports that Mercor, now valued at $10 billion, has been offering to pay people for prior work that may be owned by their employers, creating potential intellectual property exposure for the company and its customers.

The incident highlights vendor concentration risk as AI labs increasingly rely on third-party data providers rather than generating training data internally. Mercor has positioned itself as critical infrastructure for AI training data, and a security compromise at a single vendor can simultaneously impact multiple competitors. The company's business model of acquiring existing work rather than commissioning new data generation also creates legal ambiguity around ownership rights.