Play on Google Cloud — usage, spend, credit history

The setup

ProductPlay — a workspace where an AI co-worker builds the tool you describe. Live, paying customers.
Engineering team2 people. No other developers, no contractors.
Who writes the codeSupervised agents. Most commits are machine-authored, human-reviewed.
Why the bill scalesEvery agent commit is compiled, tested and stored before it counts. More agents, more build minutes.
CloudGoogle Cloud only, since day one. Nothing on AWS or Azure.

Commits · last 30 days

Play2 engineers

10,295

Elastic3,700 seats

7,410

HashiCorp2,400 seats

6,995

Grafana1,400 seats

6,192

SUSE2,500 seats

4,036

Public commit history, 28 Aug – 27 Sep 2026. Same counting rules on every row: merge commits and bot accounts removed. Seat counts are public headcount.

What runs on Google Cloud

Compute EngineSelf-hosted GitHub runner fleet, managed by GARM. Spot instances first; evictions absorbed on our side.
ARM zoneDedicated, for ARM builds and tests.
Cloud StorageBuild artifacts, logs, CI watchdog markers.
Per pull request17 jobs across a 20,608-file repository. ~15 minutes to green.
Last 30 days1,563 merged pull requests.
Not on Google yetModel inference inside the product. Currently third-party APIs.

Spend

All Google Cloud credits at 0% remaining — Every credit line at 0% remaining or expired.

90-day spend$31,560.37
Run rate~$11,000 / month, rising as we add agents
Credits applied$0.00 — first period we've paid the full amount

Credit history with Google

GFS Cloud Program — Start$2,341 · spent

GFS Cloud Program — Ecosystem Partner$26,922 · spent

GenAI App Builder + free trials$1,880 · expired

Scale tier · AI-first tiernever received

Two grants, both fully consumed. No request since. Co-marketing with Google Cloud: done once, happy to repeat with this year's numbers.

What we'd like to add — with you

Build capacityRaise the agent count instead of rationing CI. Expect the bill to grow with it.
Gemini on VertexMove the product's inference onto Google. Decision is live now, not later.
Co-marketingRepeat the last one, with the numbers above in it.

Fits: Scale tier + AI-first tier

$200K + $150K over 2 years · neither previously granted

Preference is one cloud deeper rather than three in parallel. Happy to walk through any number on this page.