Fermi estimate of Google's query throughput in queries per second

Estimate how many queries Google answers each second. Walk through how you structure the estimate, label every assumption you make, and finish with a number, a plausible range and how confident you are in it. ```hint Get to a day first Decide whether you will build up from the people who search and how often they do it, or from some measure of Google's capacity, and reach a daily total before converting to seconds; daily habits are much easier to reason about than per-second rates. ``` ### Clarifying Questions - Does "queries" mean only searches on Google Search, or also searches inside YouTube, Maps, Gmail and the voice assistant? - Should the answer be the average rate across a day, or the peak rate the system has to handle? - Do autocomplete suggestion requests and automated traffic from bots and scrapers count, or only searches a person submits? - Is the scope worldwide, or a single country? ### What a Strong Answer Covers - A scoped definition of what counts as a query (which Google products, people versus automated traffic) and of whether the answer is an average or a peak, settled before any arithmetic - A top-line equation that links search demand, or Google's capacity, to queries per second - Labeled, round assumptions, each with a one-line rationale grounded in how people actually use search - Arithmetic that is correct and unit-consistent, including the conversion from a day to seconds - A sanity check from a second angle, a final range, and how the busiest hour compares with the average ### Follow-up Questions - How far above your average would the busiest hour be, and what drives the gap? - Which assumption moves your answer the most, and what data would you use to tighten it? - How would the number change if autocomplete suggestion requests, or searches inside YouTube and Maps, were counted?