Size the Yearly Server, Disk and Accelerator Spend Behind One Google Photos Account

Quick Overview

A product manager estimation question that asks for the yearly hardware cost of supporting one Google Photos user, across storage, compute, machine-learning accelerators and networking. It tests defining the unit, structuring the equation, labeling assumptions, doing the arithmetic and sanity-checking the result.

Size the Yearly Server, Disk and Accelerator Spend Behind One Google Photos Account

Company: Google

Role: Product Manager

Category: Product / Decision Making

Difficulty: medium

Interview Round: Other

You are a product manager at Google. Estimate the hardware cost of supporting one Google Photos user: the machines and equipment Google needs so that a person can keep their photos and videos in the service and use its features. Explain how you would structure the estimate, state each assumption, work through the numbers and check whether your result is reasonable. ```hint Follow the bytes and the work Split the cost into what a user keeps on disk and the computing their uploads and browsing cause. Then ask how many bytes of physical disk one stored byte really occupies once you account for safety copies and spare space. ``` ### Clarifying Questions - Does "hardware" cover only servers, drives, accelerators and network gear, or also power, cooling and data-center buildings? - Should the cost be per user per year, per month, or over the lifetime of a user's library? - Is the target the average across everyone who uses Photos, including free accounts with small compressed libraries, or a typical paying subscriber with a large one? - Should machine-learning features such as search, face grouping and AI editing be included, or only storing and serving media? ### What a Strong Answer Covers - A defined unit (cost per user per period) and an explicit scope for what counts as hardware. - An equation that separates storage, general compute, machine-learning accelerators and networking. - Labeled round-number assumptions for library size, file sizes, redundancy and spare capacity, hardware prices, service life and utilization. - Arithmetic that is easy to follow, plus a sanity check against an independent reference point such as a top-down total or a consumer storage price. - Sensitivity: which assumptions dominate, how light and heavy users differ, and what the number implies for product decisions. ### Follow-up Questions - Which single assumption moves your answer the most, and how would you narrow it down with data Google already has? - How would your estimate change if every photo were kept at original quality instead of as a compressed copy? - Based on your number, would you change how much free storage Google Photos users get? What else would you need to know first?

Overview: A product manager estimation question that asks for the yearly hardware cost of supporting one Google Photos user, across storage, compute, machine-learning accelerators and networking. It tests defining the unit, structuring the equation, labeling assumptions, doing the arithmetic and sanity-checking the result.

Google logo
Google
Aug 16, 2025
mediumProduct ManagerOtherProduct / Decision Making
0
0

You are a product manager at Google. Estimate the hardware cost of supporting one Google Photos user: the machines and equipment Google needs so that a person can keep their photos and videos in the service and use its features. Explain how you would structure the estimate, state each assumption, work through the numbers and check whether your result is reasonable.

Clarifying Questions Guidance

  • Does "hardware" cover only servers, drives, accelerators and network gear, or also power, cooling and data-center buildings?
  • Should the cost be per user per year, per month, or over the lifetime of a user's library?
  • Is the target the average across everyone who uses Photos, including free accounts with small compressed libraries, or a typical paying subscriber with a large one?
  • Should machine-learning features such as search, face grouping and AI editing be included, or only storing and serving media?

What a Strong Answer Covers Guidance

  • A defined unit (cost per user per period) and an explicit scope for what counts as hardware.
  • An equation that separates storage, general compute, machine-learning accelerators and networking.
  • Labeled round-number assumptions for library size, file sizes, redundancy and spare capacity, hardware prices, service life and utilization.
  • Arithmetic that is easy to follow, plus a sanity check against an independent reference point such as a top-down total or a consumer storage price.
  • Sensitivity: which assumptions dominate, how light and heavy users differ, and what the number implies for product decisions.

Follow-up Questions Guidance

  • Which single assumption moves your answer the most, and how would you narrow it down with data Google already has?
  • How would your estimate change if every photo were kept at original quality instead of as a compressed copy?
  • Based on your number, would you change how much free storage Google Photos users get? What else would you need to know first?