The Hyperscaler Consensus

This week, three things happened in pharma AI. McKinsey published data showing that nearly half of healthcare organisations are pursuing hyperscaler partnerships as their primary AI strategy. McKinsey also announced its own partnership with Google Cloud to sell enterprise AI transformation. And Merck signed a deal worth up to $1 billion with Google Cloud to build an agentic platform across its entire value chain.

This week, three things happened in pharma AI. McKinsey published data showing that nearly half of healthcare organisations are pursuing hyperscaler partnerships as their primary AI strategy. McKinsey also announced its own partnership with Google Cloud to sell enterprise AI transformation. And Merck signed a deal worth up to $1 billion with Google Cloud to build an agentic platform across its entire value chain.

Three separate announcements. One pattern.

The hyperscaler is now the default answer to the pharma AI question. When a board asks "what is our AI strategy?", the answer increasingly is: we signed with Google, Microsoft, or AWS. The announcement gets made, the press release gets issued, and the AI strategy box gets ticked.

The scale of what's being built puts this in context. Global IT spending will reach $6.31 trillion in 2026 — up 13.5% year on year — with data centre investment growing 55.8% to surpass $788 billion, according to Gartner. By 2031, hyperscalers including Google, Microsoft, and AWS will account for 67% of global data centre capacity. The infrastructure race has already been won. The hyperscalers are not bidding for pharma's business. They are building the only roads, and pharma is signing long-term leases on the tollgates.

I want to ask a different question. Who captures the value?

Hyperscalers monetise on day one. Cloud spend is a contract, not a performance agreement. Google Cloud bills Merck for compute regardless of whether the agentic platform delivers faster launch cycles or not. McKinsey charges for transformation regardless of whether the workforce actually changes how it works. The ROI question — when does pharma capture the value? — is almost never in the announcement.

McKinsey's own data shows 64% of implementing healthcare organisations report anticipated or realised positive ROI. Note the word "anticipated." ROI expectations and ROI realisation are not the same metric, and in a sector where Gartner has forecast that over 40% of agentic AI projects will be cancelled by 2027, the gap between those two numbers matters.

The Merck deal has a number in the headline: up to $1 billion. That ceiling is doing rhetorical work. It signals ambition, not commitment. The number that doesn't appear in the announcement is the value Merck expects to capture, the timeline on which it expects to capture it, and the mechanism by which Google Cloud shares in the outcome if the platform underperforms.

None of this means the Merck deal is a bad decision. It may be exactly the right call given their LOE exposure and their need to compress launch cycles before the Keytruda cliff arrives. The partnership gives them embedded engineering capacity and hyperscale infrastructure at a moment when building that in-house would take longer than they have.

But the pharma industry is signing these deals at scale, and the question of value capture is being left for someone else to answer later. The hyperscaler captures revenue now. The consultancy captures fees now. Pharma captures productivity gains — maybe — in three to five years, against KPIs that mostly don't exist yet.

That asymmetry deserves a name. And it deserves a governance response before every major pharma has signed the same deal with the same three cloud providers and the differentiation has moved entirely to who can actually change how their people work.

That is the harder problem. It is also the one nobody's announcement addresses.