The Incumbent's Playbook: Why Pharma Is Taking Seats at the AI Table
When Vas Narasimhan joined Anthropic's board last week, the announcement read like a standard corporate governance item. A pharma CEO, an AI safety company, a seat at the table. But look closer at what it actually signals — and what it says about every other major pharmaceutical company that isn't building its own AI lab from scratch.
Narasimhan, the CEO of Novartis, became the first pharmaceutical executive to hold a board seat at a major AI laboratory when Anthropic appointed him through its Long-Term Benefit Trust on April 14. The Trust, an independent governance body whose members have no financial stake in Anthropic, has now appointed a majority of the company's seven-person board. Anthropic is weighing a potential IPO, making the governance structure suddenly relevant to a much wider set of investors.
The Narasimhan appointment is not an isolated event. It is the sharpest example of a pattern that plays out every time a major platform shift arrives in healthcare: the incumbents don't build competing infrastructure. They buy a seat at whoever does.
The evidence is scattered across the past two years. Novo Nordisk announced a partnership with OpenAI on the same day as the Narasimhan appointment — drug discovery and workforce training, the companies said, without specifying timeline or scope. Emma Walmsley, the departing GSK CEO, has sat on Microsoft's board since 2020. David Ricks, Eli Lilly's CEO, sits on Adobe's board. None of these are AI labs in the pure sense that Anthropic, OpenAI, or DeepMind are. But they are all companies where AI strategy is existential, not peripheral.
Pharma knows this dance. When the internet arrived, incumbents didn't build their own internets. When mobile health emerged, they didn't launch app stores. They partnered, invested, and waited to see which startups would become infrastructure. AI is the same pattern wearing a different label. The board seat is just the most explicit version of the strategy.
For the pure-play companies — the Recursions, the Exscientias, the Insitros, the Generate:Biomedicines of the world — the Narasimhan appointment should land as a warning. When the incumbents you're supposed to be disrupting have governance ties to the AI labs you're building on top of, something structural shifts. Not through competition, but through access. A pharma CEO on an AI lab board means pharma gets a voice in priorities, visibility into roadmap decisions, and relationships with the researchers who build the models that everyone else also depends on.
The Long-Term Benefit Trust structure adds another layer. Designed to ensure Anthropic's decisions account for societal risk rather than pure shareholder return, it is the governance mechanism an AI lab uses when it wants to be taken seriously by regulators, partners, and potential public market investors. Narasimhan's appointment through it signals that Anthropic wants pharma's expertise — and pharma's trust with regulators — as it navigates what could become the most consequential IPO in the sector's recent history.
What this story is not: a demonstration that every major pharma CEO has rushed onto an AI lab board. They have not. The headline premise of the original pitch was wrong. Novo Nordisk has a partnership with OpenAI, not a board seat. Ricks sits on Adobe's board. Walmsley sits on Microsoft's. The pattern is real but thinner than the pitch suggested.
What the story is: one board appointment that tells a larger story about how an industry chooses its AI future. The short version is this — pharma is not fighting the AI transition. It is renting a seat at the table, the same way it has every time a new technological substrate arrived that it couldn't build faster or cheaper itself.
The question for the AI-native drug discovery companies, the investors who fund them, and the researchers who join them: what happens when the incumbents have already been seated, and the table is getting crowded?
That is the story. The board seat is just how you can see it happening.
