Strategic positioning: Google competes on infrastructure scale and integration; Anthropic competes on safety-first protocol capture. Google’s 515 mentions reflect its breadth—search (89.9%), Android (70.7%), cloud (11% Q4 2025)—but its AI strategy is a horizontal platform play spanning Gemini, Google Cloud TPUs, and DeepMind research. Anthropic’s 900 mentions, despite no consumer OS or search, signal outsized influence via vertical differentiation: safety as a product feature and Claude Code as a developer gateway. The gap in mention volume (900 vs 515) is deceptive—Anthropic punches above weight because its narrative is concentrated, while Google’s is diluted across many business units.

Product and ecosystem: Google’s moat is infrastructure ownership; Anthropic’s is protocol lock-in. Google offers Gemini (multimodal), Vertex AI (enterprise MLOps), and TPU v5e/v5p. Anthropic offers Claude 3.5 Sonnet (92 on key benchmarks) and Claude Code, whose 21 mentions/7d drove the causal chain to MCP v2.0. The critical divergence: Google controls the compute stack (TPU + GCP), giving it cost and latency advantages. Anthropic controls the developer workflow protocol (MCP), creating switching costs that are independent of model performance. Google wins on raw scale; Anthropic wins on developer mindshare stickiness.

Recent momentum: Google is consolidating; Anthropic is standardizing. Google’s Brain-DeepMind merger (April 2023) unified research, and its 515 mentions span hardware (Gemini TPU), cloud (Vertex), and consumer (Search). Anthropic’s 900 mentions are driven by Claude Code adoption forcing protocol evolution—MCP v2.0 is a direct response to real usage patterns. The Huawei parallel ecosystem (6 mentions) appearing in the same cluster as Nvidia/Google but disconnected from Anthropic/OpenAI suggests Anthropic is decoupling from hardware dependence, betting on protocol ubiquity rather than silicon.

The critical question: Can Google convert infrastructure ownership into developer loyalty, or will Anthropic’s protocol-first strategy make model performance secondary? Google’s risk is that its hardware moat becomes commoditized as inference costs drop; Anthropic’s risk is that MCP forks or gets absorbed by open-source alternatives. The winner will be the company that makes its advantage self-reinforcing—Google by tying TPU to Gemini to GCP, Anthropic by tying Claude Code to MCP to enterprise safety compliance. The next 12 months will reveal whether protocol or infrastructure becomes the binding constraint.