Microsoft, Alphabet, Meta and Amazon are expected to spend more than $700 billion on capital expenditures this year as they build the infrastructure for AI.
That scale demands a payoff. If AI is going to justify that outlay, it has to produce measurable productivity gains across the economy, not just more computing capacity.
Healthcare is one of the clearest places to look. U.S. healthcare spending reached $5.3 trillion in 2024, or 18% of GDP. By some estimates, close to $1 trillion goes toward administration alone.
Demand for care also continues to outpace the available clinical workforce. That makes healthcare a natural test case for whether AI can lower the cost of delivering services while freeing clinicians to serve more patients.
Major players are already moving. Microsoft and Mayo Clinic are developing a frontier AI model designed specifically for healthcare. Google Cloud and CVS Health have also entered a long-term partnership to power CVS’s new Health100 platform with Gemini.
For providers, the economics are straightforward. Reducing administrative work lowers the cost of delivering care. But the window to capture those savings may be limited, because reimbursement tends to reflect greater efficiency over time.
Medicare already incorporates productivity into payment updates across hospitals, skilled nursing facilities, home health and other settings. For 2026, CMS applied a 2.5% efficiency adjustment to the work component of certain non-time-based physician services.
That means early adopters have an incentive to move now. Providers that use AI to reduce documentation, intake or coordination costs can improve their cost structures before reimbursement catches up.
Those that wait could face the same payment pressure without having banked the savings. In that sense, AI is not just a technology decision. It is a margin decision.
The market has already voted on the first phase. Capital is flowing into chips, servers and data centers. The harder question is what businesses can do with that infrastructure once it is built.
Healthcare offers a concrete answer. A delayed referral or missing information can trigger manual follow-up, slow care and raise the risk of readmission. If AI helps a handoff happen correctly the first time, it can eliminate downstream work and cost.
That is the kind of productivity gain that matters. Time saved on documentation or scheduling is not useful if it simply shifts the burden somewhere else. It has to return capacity to the system.
Providers also need to decide what to do with that capacity. It can mean seeing more patients, easing pressure on stretched staff, or reinvesting in staff and access. Either way, the benefit depends on turning software into output.
For the broader economy, that is the real test of the AI boom. Capital rewards clear rules, but it only compounds when productivity rises. Healthcare will show whether this $700 billion bet can produce that result.
