I have been staring at Naftiko Signals for healthcare longer than I have stared at any other vertical. Ninety-three companies — payers, providers, health systems, biopharma, medtech, diagnostics, pharmacy retail — adding up to an aggregate signal score of 35,078 across 2,510 areas, 437 services, 258 tools, and 249 standards. That makes healthcare the largest single industry footprint we read at Naftiko, and the one where I think the technology, business, and politics of APIs collide hardest in public.
The headline numbers
The top of the score breakdown is unsurprising in shape and surprising in scale. Services (5,421), Data (2,719), Cloud (2,541), Operations (1,721), and Automation (1,526) lead the table, with Security (1,393) and Artificial Intelligence (1,331) right behind. Healthcare is not a sector that is dabbling in cloud-native, AI-adjacent infrastructure — it has standardized on it. The companies sitting on top of the rankings are exactly who you would expect: UnitedHealth Group leads in 15 of 20-plus scored areas (Cloud 171, Data 143, Security 93, Services 308), AstraZeneca is the consistent top-three across cloud, AI, multimodal and data, and Philips holds first place in Specifications and SaaS and second or third across nearly every foundational area. Johnson & Johnson, Abbott, Kaiser Permanente, and Medtronic form the dense mid-tier behind them.
Interoperability and the AI stack
The detections underneath those scores are where the conversation gets honest. The interoperability footprint is enormous — 214 companies reference interoperability as an area, and the patterns underneath it surface Epic, FHIR-bound EHR integrations, HL7 messaging, pharmacy benefit manager APIs, transplant registries, medical device telemetry, and a layer of WhatsApp and Microsoft Teams notifications wrapped around clinical workflows. Underneath the AI talk, the actual deployed services tell the real story: Azure Machine Learning (653), Hugging Face (451), GitHub Copilot (311), Gemini (247), Claude (191), OpenAI (186), Anthropic (156), Amazon SageMaker (125), Azure Databricks (338). That is not a sector experimenting with AI. That is a sector that has chosen a stack and is hiring against it.
The cloud picture is just as concentrated. Microsoft Azure shows up everywhere — Azure Functions, Azure Service Bus, Azure Data Factory, Azure Databricks, Entra — alongside Oracle Cloud, Snowflake, and Databricks for the analytics layer. Git, Terraform, PowerShell, PostgreSQL, Prometheus, Kubernetes, SonarQube, Semantic Kernel, Datadog, New Relic, and ServiceNow are deployed across all 31 of the deeply measured companies. The platform game in healthcare is over. What is being decided right now is what gets built on top of it.
Heavy data, thin specialization
And here is the contradiction the public footprint exposes. Healthcare scores 143 on Data and 171 on Cloud for its leader — and a 2 on Domain Specialization. The richest clinical, claims, genomic, and device datasets in any vertical, sitting on top of one of the most mature cloud and data toolchains in any vertical, and the formal fine-tuning, model registry, and domain-adaptation discipline to actually turn that data into clinical-grade AI tops out at a score of 2. Pair that with Privacy & Data Rights topping out at 6 across the whole sector — well below the Security investment that pays for keeping the data from being stolen — and you can read the strategic posture clearly: healthcare has invested in protecting data and accumulating it, but not yet in governing how that data is used once it is inside the enterprise. The Specifications score (Philips leads at 17, everyone else trails) tells the same story for interoperability — the standards conversation is loud in public, the formal API and schema work is thin in practice.
What’s next
So what’s next. Two industry-level capability recommendations I keep coming back to. First, make domain specialization a first-class capability — productionize fine-tuning pipelines and model registries against the clinical, claims, and genomic data assets you already have, before generic foundation models get deployed into workflows that carry patient safety and regulatory consequences. Second, treat API specifications and privacy engineering as the same problem — FHIR and HL7 are not enough on their own; the surface area of how AI agents read, write, and reason over PHI needs schema discipline, consent flows, and algorithmic-impact documentation built in from the start. Those are the moves that turn healthcare’s data and security strength into something durable.
Read the full Signals breakdown for the industry at https://industries.naftiko.io/signals/healthcare/PXUVl6R6WS/.
