On November 16, 2026, Vercel added per-project and per-product budgets to Spend Management, alongside the team budget that already existed. A project budget can pause production for that project alone once spend reaches the amount you set. Until now, automatic pausing only existed on the team budget, and it takes every project on the team offline at once.
The feature is available at no extra cost to Pro teams and to Enterprise teams on the Flexible Commitment plan, and only people with the Owner or Billing role can configure it. Budgets live on the new Spend Management page and in the CLI, with vercel billing budgets set --project <name> --amount <value> --pause. A product budget (Vercel's example is Fast Data Transfer) only sends alerts; it pauses nothing.
Pausing means the site is down
The documentation is plain about the effect: while production is paused, the project's websites, APIs, and functions are unavailable, and visitors get a 503 DEPLOYMENT_PAUSED error. The pause does not undo itself. Raising the budget does not bring the project back; someone has to resume each project from the dashboard, the budget's menu, or the API.
The cap is not exact either. Vercel checks spend every few minutes, and during that window the project keeps serving traffic and accruing usage. The documentation itself recommends setting the budget below the most the company is willing to pay. Deployment storage, Vercel Blob, and Container Registry images are still billed while paused, and AI Gateway and v0 usage neither stops nor counts toward a project budget.
On Pro, a project budget counts the monthly credit
This is the part most likely to surprise. On Pro, the team budget only counts usage beyond the plan's monthly credit. A project budget counts all of that project's usage, including the part the credit covers. In practice, a project can be paused while the team's invoice would not yet show any extra charge. Anyone who copies the team-budget logic onto a project will set the wrong amount.
Who should look at this now
The clearest case is agencies and teams that run several client sites or products in a single Vercel team. Before, anyone who wanted a cap that actually stopped spending had a bad choice: turn on the team pause and accept that one project with abnormal traffic could take down all the others, or settle for alerts. Now the project at risk of a spike (a campaign, a public API, an experimental environment) can be isolated without exposing the rest.
It is the reverse of what Cloudflare shipped three days earlier. With Cloudflare's spending limit for Workers and Durable Objects, the cap applies to the whole account, and a test Worker can block production in the same account. Teams on both platforms now have controls with different reach and should write that down so nobody assumes the same behavior.
What to decide
There is no deadline: everything is opt-in and nothing changes for teams that create no budgets. The decision, project by project, is whether a bigger bill is worse than the site being down. For a corporate site or a store, the answer is usually no, and the route is a budget with alerts (50%, 75%, and 100%) or a webhook, without pausing. For a prototype, a demo, or an API with no availability commitment, pausing protects the bill at little real cost.
If you turn pausing on, decide beforehand who can resume the project and how fast, because only Owner and Billing roles can do it and resuming is manual. Also check that alerts reach someone who reads them, since a pause can happen outside business hours. For teams reviewing how they pay Vercel as a whole, this decision sits next to the one about fixed CDN pricing, which also depends on the team's usage tier.
What Vercel has not spelled out yet
The documentation says spend is checked "every few minutes" without a number, which leaves open how far a project can overshoot its budget before it stops. The organization budget, which adds up several teams, is limited to Enterprise on Flexible Commitment and, according to the announcement, to customers in the Organizations private beta, with no general availability date. And since AI Gateway usage does not count toward project budgets, anyone running AI per project needs a second control: AI Gateway's own budget.
If your company is weighing something similar, reach me on LinkedIn or WhatsApp.