Back-of-envelope: serving 1B AI Mode queries on Gemini costs only ~$4M
gaganghotra_ · x · 2026-11-13
Pushing back on SEO-conference claims that Google's AI Mode is expensive to run, the author estimates that with 8 input and 533 output tokens per query (6-word query, 400-word answer), serving 1 billion queries at public Gemini API prices costs only about $4M net.
A chart compares model tiers: Gemini 3.8 Flash $2.0M, Gemini 4 Hydrogen $4.0M, and Gemini 3.5 Flash $4.8M for the same 1B answers. Output tokens account for 99.7% of the cost. At Google's scale of billions of queries, inference cost is effectively pocket change.
More from Venture
- Ex-Moonshot AI researcher startup uses AI to automate chip formal verification — teortaxesTex · 2026-11-13
- Anthropic's quarterly revenue grew 45x in 18 months, hitting $17B by Q3 '26 — Dr_Singularity · 2026-11-13
- Israel's VCs See Opportunity for Challengers in Nvidia's Turmoil — TechCrunch AI · 2026-11-13
- Indie hacker advice: you only need a $3-5k/month niche, not a huge market — gefei55 · 2026-11-13
- NVIDIA's six main suppliers grew 64% YoY last quarter, fastest in a year — tengyanAI · 2026-11-13
- claude-trading-skills hits 3.8k stars: Claude Code skills for equity traders — tradermonty · 2026-11-13