Vibe coding binge leads to $800 Vercel surprise bill
Matthew Berman received an unexpected $800 Vercel bill after two weeks of AI-assisted "vibe coding," caused by defaulting to the highest-cost build machine, concurrent duplicate builds, and slow build times — all of which he fixed after community feedback.
Score breakdown
Developers using AI coding assistants to ship fast should audit cloud deployment defaults and build configurations before costs spiral — AI tools optimize for speed, not cost efficiency.
- 01Berman received an $800 Vercel bill after just two weeks of AI-assisted vibe coding while building a project called Journey Kits.
- 02Vercel defaulted to its 'turbo build machine,' the most expensive option, charging 12 cents per build minute.
- 03The Elastic tier Berman switched to starts at 3/10 of a penny per build minute — a drastically lower rate.
Matthew Berman describes how a highly productive stretch of AI-assisted "vibe coding" — during which he shipped multiple products — ended with a jarring $800 Vercel bill arriving after just two weeks. He attributes the problem to moving too fast and trusting his AI coding assistant's deployment recommendations without scrutinizing the underlying configurations. Vercel's defaults placed him on the "turbo build machine," its most expensive option at 12 cents per build minute, compared to the Elastic tier which starts at 3/10 of a penny per minute. On top of that, Vercel's default setting of running all builds immediately meant that his habit of deploying dozens of times per day — often making a small change and redeploying before the previous build finished — resulted in multiple concurrent builds charging simultaneously for essentially the same output.
After posting about the bill on X, Berman received community feedback, including input from a user named Theo who flagged that something was making his builds unusually slow.
After posting about the bill on X, Berman received community feedback, including input from a user named Theo who flagged that something was making his builds unusually slow. Slow builds compounded the cost problem since Vercel charges per build minute. Acting on the suggestions, Berman switched from the turbo machine to the Elastic tier, disabled on-demand concurrent builds so they run sequentially (allowing him to cancel or queue redundant deploys), and optimized his build process. Build times dropped from over three minutes — sometimes four — down to roughly one minute per build, with further room for improvement. One community member also suggested offloading the build step to GitHub hooks and using Vercel only for deployment. The result: his weekly Vercel costs fell from hundreds of dollars to just a couple of dollars.
Key facts
- 01Berman received an $800 Vercel bill after just two weeks of AI-assisted vibe coding while building a project called Journey Kits.
- 02Vercel defaulted to its 'turbo build machine,' the most expensive option, charging 12 cents per build minute.
- 03The Elastic tier Berman switched to starts at 3/10 of a penny per build minute — a drastically lower rate.
- 04Rapid AI-assisted development led to deploying dozens of times per day, often triggering duplicate concurrent builds all charged simultaneously.
- 05Berman disabled Vercel's default 'run all builds immediately' setting in favor of sequential builds, allowing cancellation of redundant deploys.