Strategic positioning — Google is playing a platform arbitrage play, while OpenAI is playing an application-layer capture play. Google’s 515 mentions cluster around infrastructure (TPU v5e, Gemini API, Google Cloud’s 11% share) and ecosystem (Android, Search, YouTube). It treats AI as a horizontal layer that strengthens existing moats. OpenAI’s 653 mentions center on model frontier (o3-mini, GPT-4o) and protocol lock-in (Claude Code, MCP). OpenAI is not defending a search monopoly — it’s building a new one: the developer’s runtime. The critical asymmetry: Google can afford to lose model race because it owns distribution; OpenAI cannot.

Product and ecosystem — Google’s moat is infrastructure bundling: Gemini 2.0 Pro, Vertex AI, and Colab Enterprise create a closed-loop for enterprises already on GCP. OpenAI’s moat is developer workflow capture: Claude Code’s 21 mentions/week and MCP v2.0’s imminent release signal that OpenAI is winning the agent-to-tool protocol layer. Google’s response (A2A protocol) is defensive and late. The real battle isn’t Gemini vs GPT-4o — it’s Vertex vs. ChatGPT API + MCP. Google owns the cloud rack; OpenAI owns the developer’s keyboard.

Recent momentum — OpenAI is accelerating protocol lock-in faster than Google can build cloud-native AI. MCP v2.0 hardening (causal: Claude Code adoption) creates a network effect that Google’s open-source A2A can’t match because Google lacks a dominant agent runtime. Meanwhile, Huawei’s 6 mentions in a parallel ecosystem (decoupled from Nvidia) signals that Google’s hardware advantage (TPU) is being challenged by a third pole — and OpenAI is agnostic to that fight. Google’s 5 Meta mentions suggest it’s watching the open-source threat, not OpenAI.

The critical question — Can Google convert its infrastructure distribution advantage into developer mindshare before OpenAI’s protocol layer becomes the default runtime for agents? If MCP becomes the TCP/IP of AI agents, OpenAI controls the routing layer — and Google becomes just another cloud provider. If Google bundles Gemini into Android and Search at scale, it wins the consumer layer while OpenAI wins the professional layer. The war is not models — it’s who defines the invocation standard. Right now, OpenAI is winning that battle by 2x mention volume and a protocol that’s already in production.