In Healthcare AI, the Pipes Matter More Than the Models

4 min read.

Author

Midas AI
August 14, 2025

HFMA’s latest survey and Oracle’s AI first EHR both point to the same reality. Control of data, workflows, and cloud delivery will decide who captures value from AI in healthcare.

If you want to see where the AI race is actually being won in healthcare, look past model specs and into the supply chain. A new Healthcare Financial Management Association report finds health systems moving fast to install AI governance and vendor guardrails, even as most lack the resources to stand up AI on their own.

Days later, Oracle unveiled a cloud built AI and voice driven EHR that embeds its clinical AI agent directly into clinician workflows. The throughline is clear. In healthcare, distribution through the EHR, access to patient data under governed policies, and cloud scale infrastructure are emerging as the real competitive levers.

AI Adoption Realities and Procurement Pressures

Only 18% of health systems report having mature AI programs, and three quarters of those programs sit in organizations with more than 1 billion dollars in annual net patient revenue. Even among the mature cohort, only half say they have sufficient resources to implement AI tools. Across the broader field, more than 80% say they do not have the resources to identify, select, and implement AI solutions.

Governance is catching up quickly, with 70% of chief financial officer respondents reporting any type of AI governance process in place, up from 41% in 2024. Yet more than a third of health systems already investing in AI have no formal data policy for AI use. As the HFMA report puts it, internal governance and thoughtful AI strategy are critical, and health systems must ensure that skills and resources exist to identify opportunities, vet vendors, implement solutions, and evaluate long term ROI.

Those constraints shape procurement. Three quarters of respondents say they would prefer to work with vendors integrated with their current EHR system, and 11% have no plans to invest in AI until a solution is available through their EHR. Almost 80% say one of their existing vendors, or an AI company partnered with an existing vendor, would have an advantage in winning an AI pilot, and nearly as many indicate greater comfort sharing data with an existing vendor.

CFOs and CIOs are most often the decision makers in weighing a vendor’s AI offerings. Cost reduction is the most frequently cited rationale for AI investments, but only 39% of CFOs believe the investments will result in an overall cost reduction. In other words, distribution and trust through existing EHRs and vendor relationships matter as much as, if not more than, any model benchmark.

Oracle’s New EHR and the Competitive Shift

Oracle Health launched a cloud native platform built on Oracle Cloud Infrastructure, separate from the legacy Cerner stack, with embedded clinical AI agents that combine generative AI, clinical intelligence, multimodal voice, and screen driven assistance. The company describes the agent as contextual and conversational, with voice activated navigation and search to surface items like recent lab results and current medications.

Oracle designed a semantic AI foundation as an open system so customers can extend Oracle’s agents, build their own, or integrate third party models, and the stack is positioned for rapid deployment with enterprise grade performance, scalability, and efficiency. Seema Verma, who leads Oracle Health and Life Sciences, said Oracle chose to build from the ground up rather than bolt features onto antiquated technology, and pitched the platform as freeing providers from technical baggage so they can focus on caring and preventing illness. The initial release targets U.S. ambulatory providers pending regulatory approvals, with acute care functionality planned for 2026.

The timing and market dynamics underline how the channel is consolidating. Oracle’s share of the U.S. acute care hospital EHR market declined to 22.9% in 2024 as the company experienced a net loss of 74 hospitals and 17,232 beds, while Epic added 176 multispecialty hospitals and now holds 42.3% of the market.

Oracle’s rollout lands just before Epic’s annual user meeting, where Epic is reportedly set to announce an AI powered ambient clinical documentation tool. The competitive field is not about who builds a better model in isolation. It is about who embeds AI into the clinician’s daily screen, pipelines the data securely under a hospital’s governance, and operates at cloud scale.

This is also why capital alone does not guarantee deployment. AI enabled startups captured 62% of digital health venture dollars, nearly 4 billion dollars, in the first half of 2025. Yet the HFMA survey shows incumbents and EHR integrated offerings will often win the pilot and the data share.

Governance groups, where present, are tasked with setting AI data policy, vetting vendors with AI capabilities, and identifying AI vendors for internal initiatives. The White House’s recent AI Action Plan adds a national policy backdrop, but inside hospitals the practical gatekeepers remain EHR integration, existing vendor relationships, and the ability to deliver within regulated workflows.

In healthcare, compute supply chains are the strategy. Winning means owning or partnering into the EHR workflow, aligning with hospital data policies, and delivering through cloud infrastructure that clinicians can actually use. For founders and operators, the shortest path from model to impact runs through the pipes.