I went straight into storage costs because that felt obvious, but then realized I was ignoring compute, bandwidth, ML inference for things like face grouping, and the support infrastructure underneath all of it.
Break down the cost into storage, compute, and network components, then estimate each using publicly known Google infrastructure metrics and reasonable assumptions. Structure your answer top-down, starting with total users and average storage per user, then apply unit costs for storage, serving, and processing. Conclude with a rough annual cost and note key uncertainties.
Pro tip: Show product sense by linking cost drivers to product decisions—e.g., how offering unlimited free storage (historically) impacts cost, and how features like auto-backup and AI editing increase compute costs. This demonstrates you think about profitability and trade-offs, not just arithmetic.
Clarify what 'run Google Photos' includes: storage, serving, processing, and maintenance. Estimate monthly active users (MAU) and total users, using known Google Photos user numbers (e.g., 1B+ MAU).
Assume average photos/videos per user and average file size to compute total petabytes stored. Multiply by Google's cost per GB for storage (e.g., $0.02/GB/month for cold storage, higher for hot).
Account for serving images (bandwidth), running AI features (face grouping, search), and data processing. Use rough estimates for compute per user or per photo, and apply Google's internal compute costs.
Add storage, compute, and network costs to get monthly cost, then multiply by 12 for annual. Sanity-check against Google's overall cloud revenue or known data center costs.
Highlight major assumptions (e.g., average storage per user, cost per GB) and how changes affect the estimate. Relate to product strategy, such as tiered storage or ad-supported models.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.