OpenAI is betting that medical data becomes the moat. Health in ChatGPT, now rolling out to eligible U.S. users, lets people securely connect medical records and Apple Health data directly to the platform, creating a personalized health insights layer that competitors cannot easily replicate. This move targets a core vulnerability of ChatGPT's API business: without proprietary data or switching costs, customers remain transactional. Medical records change that calculus. A user who connects years of health data to ChatGPT faces friction migrating to Claude or Gemini. OpenAI is transforming ChatGPT from a stateless chat interface into a persistent health knowledge system. The timing matters. Google and Microsoft have pursued healthcare AI partnerships, but neither has moved decisively into the consumer-facing medical data integration space. OpenAI is exploiting that gap before competitors consolidate healthcare relationships.
Simultaneously, OpenAI announced Project Camellia in Effingham County, Georgia—a sprawling infrastructure commitment framed around 'responsible energy' and community investment but transparently about securing reliable power for compute-intensive training and inference. The Georgia Power agreement guarantees dedicated capacity for a major OpenAI facility, removing dependency on cloud providers like AWS and Azure. This signals OpenAI recognizes the API model's vulnerability: margins compress when infrastructure costs remain hostage to third-party providers. By owning power relationships and building dedicated compute centers, OpenAI controls cost structure and availability. Notably, OpenAI also introduced Presence, an enterprise AI agent platform deploying voice and chat agents for customer workflows—another non-API product. One concrete use case: a financial services firm replaces human customer service reps with a Presence voice agent handling account inquiries and complaints 24/7, measurably reducing labor costs while improving availability.
These moves reveal OpenAI's underlying thesis: the API commodity business cannot sustain $80 billion valuations or fund frontier model training. Instead, OpenAI is building defensible products around data (Health), infrastructure ownership (Georgia), and agent deployment (Presence). The question lurking beneath all three: is OpenAI admitting the foundational model API market commoditizes faster than expected, forcing a pivot toward sticky applications and enterprise services? If so, the competitive dynamics shift dramatically. OpenAI stops racing on model capability and starts racing on moat—data, inference cost, and verticalized AI agents. That's a fundamentally different business than the API-first narrative that defined the past two years.