Ramp data: Google's cached-token share is less than half its peers', signaling agentic AI gap
Sufficient-Bat-9880 · reddit · 2026-11-04
Ramp AI Index spend data compares token usage mix across OpenAI, Anthropic, Cursor and Google. Anthropic and Cursor show near-total cached-token shares, while Google's cached share is less than half its peers'. Since heavy context reuse signals agentic workflows, the chart suggests Google's models lag in agentic AI adoption, with OpenAI's cached share climbing from 30% to 80% over the period.
More from Infra
- SurrealDB 4 launches with SurrealDB Earth: serverless agent database on Cloudflare — ritakozlov · 2026-11-04
- The hidden latency inside your database: PlanetScale on network-attached storage costs — DanielLockyer · 2026-11-04
- Binance discloses $6B+ heist, largest in history, warns of AI-accelerated social engineering — evilsocket · 2026-11-04
- US now spends twice as much building data centers as offices: $96.5B vs $46B — ohlennart · 2026-11-04
- Jensen Huang and DHH discuss inference and open-weight models at Fireworks Forge 2026 — gethackteam · 2026-11-04
- AIOps platform cuts agent token costs 30%+ with Pydantic's open-source sandbox Monty — samuelcolvin · 2026-11-04