Strategic positioning: safety brand vs. scale narrative. Anthropic has successfully weaponized its Public Benefit Corporation status and "Constitutional AI" framing into a procurement advantage, particularly in regulated industries and enterprise security reviews. OpenAI, despite its nonprofit parent, has positioned itself as the AGI frontier lab — the entity willing to move fast, ship imperfect products, and let usage data define safety. This is not cosmetic. Anthropic's safety posture converts directly into enterprise procurement wins (the 92 on Claude 3.5 Sonnet's benchmark card was accompanied by a security-hardened API), while OpenAI's positioning converts into consumer mindshare and developer defaultism. The 916 vs. 668 mention gap in your data reflects this: Anthropic generates more *analyst and enterprise* discussion because its differentiation requires explanation; OpenAI's dominance is assumed and thus less discussed.

Product and ecosystem: MCP is the quiet moat. The content performance data showing "MCP" and "async agents" trending for DeepMind and OpenAI is telling — they're playing catch-up to a protocol Anthropic open-sourced. Model Context Protocol is Anthropic's most strategically significant asset, not Claude itself. By standardizing the agent-tool interface, Anthropic has made itself the interoperability layer for the agentic era, forcing OpenAI and Google to either adopt or fork. OpenAI's response — proprietary agent APIs and a closed tool ecosystem — is a defensive move that fragments the market. OpenAI retains the developer mindshare advantage via GPT-5.6's consumer familiarity, but Anthropic's MCP adoption in enterprise stacks is a compounding moat that model benchmarks don't capture.

Recent momentum signals divergence. OpenAI's GPT-5.6 "Sol" matching Fable 5 at 1/ cost is a direct assault on Anthropic's enterprise price-performance argument. Meanwhile, Google's DeepMind async agents and Huawei's parallel infrastructure ecosystem (decoupled from Nvidia) suggest the market is fragmenting into two axes: model capability vs. infrastructure control. Anthropic's absence from the Huawei cluster is notable — it remains firmly Nvidia-bound, while OpenAI's compute deals give it more flexibility. The trend data showing Anthropic unconnected to Claude Code in the same cluster as Nvidia/Google but not Anthropic/OpenAI suggests Anthropic's developer tooling is underweighted relative to its model brand.

The critical question: can Anthropic survive model commoditization? GPT-5.6's cost-performance parity means Claude's safety premium will erode. Anthropic's bet is that MCP becomes the TCP/IP of agents, making it the protocol owner while OpenAI becomes just another model vendor. But protocols are commoditized by adoption, and OpenAI's distribution — ChatGPT's hundreds of millions of users — can fork or ignore MCP at will. The rivalry is no longer about model quality; it's about who controls the agentic default standard, and Anthropic's MCP is currently the only credible open alternative to OpenAI's closed loop. If MCP wins, Anthropic wins. If OpenAI's proprietary agent stack wins by sheer user volume, Anthropic becomes a niche safety vendor. That asymmetry — protocol ownership vs. distribution scale — is the defining tension.