Estimation: sizing Google's total computer fleet

Quick Overview

A product manager estimation exercise that asks how many computers Google owns. It tests defining what counts as a computer and as owned, splitting the fleet into groups with different drivers, making labeled round assumptions, cross-checking the largest group from an independent angle, and giving a defensible range.

Estimation: sizing Google's total computer fleet

Company: Google

Role: Product Manager

Category: Product / Decision Making

Difficulty: medium

Interview Round: Other

Estimate how many computers Google owns. Show how you break the problem down, label each assumption, and finish with a figure, a range and the assumption your answer depends on most. ```hint Split by driver Divide the fleet into groups whose size is driven by different things, and for each group find a quantity you can reason about with some confidence instead of guessing its count directly. ``` ### Clarifying Questions - Does "computer" mean the servers in Google's data centers, the laptops and desktops its workforce uses, or both, and do phones, tablets and test devices count? - Should servers that Google Cloud rents to outside customers count alongside the ones running Google's own products? - Is the question about Google alone or all of Alphabet, and does hardware sitting in leased or partner facilities count as owned? - Is an order of magnitude enough, or do you want the answer within a factor of two? ### What a Strong Answer Covers - A settled definition of what counts as a computer and as owned, agreed before any arithmetic - Separate estimates for groups with different drivers, each with its own top-line equation - Labeled, round assumptions with one-line rationales, and arithmetic that is correct and unit-consistent - A cross-check of the largest group from an independent angle - A final range, which group dominates the total, and the input that moves it most ### Follow-up Questions - Which single assumption moves your total the most, and what public information would you use to narrow it? - How would your answer change if the question were about the computing capacity Google uses, including capacity it rents from others, rather than the machines it owns? - If a growing share of spending goes to AI accelerators, how would you expect the number of machines and the fleet's mix to change?

Overview: A product manager estimation exercise that asks how many computers Google owns. It tests defining what counts as a computer and as owned, splitting the fleet into groups with different drivers, making labeled round assumptions, cross-checking the largest group from an independent angle, and giving a defensible range.

Google logo
Google
Nov 6, 2023
mediumProduct ManagerOtherProduct / Decision Making
0
0

Estimate how many computers Google owns. Show how you break the problem down, label each assumption, and finish with a figure, a range and the assumption your answer depends on most.

Clarifying Questions Guidance

  • Does "computer" mean the servers in Google's data centers, the laptops and desktops its workforce uses, or both, and do phones, tablets and test devices count?
  • Should servers that Google Cloud rents to outside customers count alongside the ones running Google's own products?
  • Is the question about Google alone or all of Alphabet, and does hardware sitting in leased or partner facilities count as owned?
  • Is an order of magnitude enough, or do you want the answer within a factor of two?

What a Strong Answer Covers Guidance

  • A settled definition of what counts as a computer and as owned, agreed before any arithmetic
  • Separate estimates for groups with different drivers, each with its own top-line equation
  • Labeled, round assumptions with one-line rationales, and arithmetic that is correct and unit-consistent
  • A cross-check of the largest group from an independent angle
  • A final range, which group dominates the total, and the input that moves it most

Follow-up Questions Guidance

  • Which single assumption moves your total the most, and what public information would you use to narrow it?
  • How would your answer change if the question were about the computing capacity Google uses, including capacity it rents from others, rather than the machines it owns?
  • If a growing share of spending goes to AI accelerators, how would you expect the number of machines and the fleet's mix to change?