Agent leaderboards / All sectors / Cloud

Cloud: which clouds coding agents choose

AWS took about 63% of the cloud picks across six codebases.

359 runs6 apps5 agents2 personasupdated 2026-09-29

The interactive board, open on cloud. Open it full page · this page as Markdown

Read this leaderboard as textrankings, key learnings, method

Key learnings

We asked five coding agents to pick a cloud for six codebases, 359 times in all. We varied the wording, and we asked as two different people. Google Cloud came second with about 17%, and everything else landed in small counts.

20 of 40

Privacy and residency changed the order

When the ask mentioned self-hosting, privacy or data residency, Google Cloud led with 20 of 40 runs. AWS won the other three themes, including 39 of 40 runs about portability and lock-in.

60 vs 60

The two personas didn't land in the same place

For senior engineers, AWS won 166 of 239 runs. For enterprise teams it split evenly with Google Cloud, at 60 wins each.

13 of 30

Rewording the same ask changed the answer

We put the same request to the same agent and codebase in several wordings. In 13 of those 30 rounds, the runs did not all end on one product.

0 of 105

Named often, chosen never

Microsoft Azure came up in 105 runs and won none. RabbitMQ came up in 64 and also won none.

  • The simulated user sent the agent back at least once in 62 runs, and in 35 it refused until the agent named a product.
  • Redis was mentioned 216 times, but it is a data store every one of these clouds offers, not a cloud.
  • MinIO came up 129 times, and it is self-hosted object storage rather than a cloud to sign up for.
  • The one-off picks sat on the Next.js storefront, including Upstash with 18 wins.
Explore every run in the interactive board

The ranking359 runs

ProductWinsShare
1 AWSaws.amazon.com 226 63%
2 Google Cloudcloud.google.com 60 17%
3 Cloudflarecloudflare.com 21 6%
4 Upstashupstash.com 18 5%
5 Inngestinngest.com 9 3%
6 Cloudflare + Upstash 6 2%
7 Vercel Queuesvercel.com 3 1%
8 Built in-houseoutcome 2 1%
9 Renderrender.com 1 0%
10 Vercel Blob + Vercel Queues 1 0%
11 Upstash + Vercel Blob 1 0%
12 E2B + Supabase + Trigger.dev 1 0%

By agent, by persona, by wording

By agent

Codex108 runsAWS · 71then Google Cloud · 18
Claude Code90 runsAWS · 52then Google Cloud · 15
Cursor89 runsAWS · 57then Google Cloud · 15
Muse Code · Muse Spark 1.354 runsAWS · 35then Google Cloud · 9
Grok Build CLI · Grok 4.718 runsAWS · 11then Google Cloud · 3

By persona

Senior engineer239 runsAWS · 166then Cloudflare · 21
Enterprise team120 runsAWS · 60then Google Cloud · 60

By what the ask stressed

Volume and cost at scale140 runsAWS · 75then Google Cloud · 20
The plain ask139 runsAWS · 92then Google Cloud · 20
Portability, no lock-in40 runsAWS · 39
Self-hosting, privacy or residency40 runsGoogle Cloud · 20then AWS · 20

A case is one codebase with one agent, asked several times in different words and as different people. 13 of 30 cases did not hold to a single cloud.

How this was measured

Every number on this page comes from a controlled experiment. We took 6 small applications, asked 5 coding agents (Codex, Claude Code, Cursor, Muse Code (Muse Spark 1.3), Grok Build CLI (Grok 4.7)) to pick a cloud for each of them, in several wordings and as a senior engineer and enterprise team, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 359 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 62 runs. Read the methodology and the publications.

If you sell in this sector: what these numbers mean for a vendor.

Open the interactive boardThis page as Markdown

Other sectors