It was a Thursday night and a client's Next.js site had just gone into what I can only describe as a timeout spiral. Serverless functions hanging at 9.8 seconds, users getting blank screens, and me frantically checking the Vercel dashboard at midnight trying to figure out if we'd actually hit a plan limit or if the code was just terrible. (It was both, honestly.) That project was on the Hobby plan. It probably shouldn't have been.
I've deployed well over 12,000 sites across Seahawk Media over the years. Vercel is part of our stack constantly, for Next.js work especially, nothing else comes close for developer experience. But the Hobby plan has sharp edges that are easy to miss until a project bites you. So here's my actual read on where those edges are and when they start to matter.
What the Hobby Plan Actually Gives You
Let's be precise. As of 2024, Vercel's Hobby plan gives you:
- 100 GB bandwidth per month
- 6,000 build minutes per month
- Serverless function execution timeout of 10 seconds
- 100 serverless function invocations per day... wait, no. That limit doesn't exist. But there's a 100 GB-hours of serverless function execution per month cap
- Edge function execution of 500,000 invocations per month
- 1 concurrent build at a time
- Deployments limited to personal accounts only (no teams)
- No commercial use allowed per Vercel's terms
That last one trips people up more than any technical limit. The Hobby plan is explicitly for personal, non-commercial projects. If you're building for a paying client or running any kind of business on it, you're already in violation of the terms of service. I've seen freelancers do this for years and it's a liability waiting to happen.
The 10-Second Function Timeout
This is the one that gets people. Ten seconds sounds like a lot until you're making a third-party API call that takes 4 seconds, running some database logic that adds another 3, and then doing something with the response. Suddenly you're at 9.2 seconds with no margin.
On the Pro plan that ceiling jumps to 60 seconds (and up to 900 seconds with configuration on Enterprise). That's not a minor difference. That's the difference between a working integration and a product that's fundamentally broken for certain use cases.
Build Minutes: Where They Disappear Faster Than You'd Think
6,000 build minutes per month sounds generous. And for a single personal project it probably is. The problem is that number erodes faster than expected once you're iterating seriously.
Back in 2021 I was running a side project, a content aggregator built on Next.js with about 400 static pages. I had ISR set up but I was also doing frequent full rebuilds while experimenting with the data layer. Burned through roughly 4,200 build minutes in three weeks. Not because builds were slow, but because I was triggering them constantly.
Each build on a medium-complexity Next.js site can run 4-8 minutes easily. If you're pushing to main 15 times a day during active development (which is normal for me), that's 60-120 build minutes gone in a single day. Do that for a week and you've chewed through half your monthly allocation.
There's no rollover. Minutes don't carry over month to month. And when you hit zero, your deployments stop. Full stop.
How to Slow the Burn
A few things I do on Hobby plan projects to preserve build minutes:
- Push to a non-production branch for experimental changes. Only merge to main when something is actually ready.
- Use
vercel --prebuiltwith cached outputs where the build artefacts haven't changed meaningfully. - Set up Vercel's ignored build step to skip builds when only certain files change (like readme updates or non-code commits).
- Batch commits. Instead of pushing 6 small fixes separately, stage them and push once.
None of this is revelatory. But in practice most devs on Hobby plans just push freely and wonder why they're out of minutes by the 20th of the month.
Bandwidth: Usually Fine, Occasionally Not
100 GB bandwidth per month is where the Hobby plan is actually pretty reasonable. For a personal project, a portfolio, a small blog, even a modest SaaS side project with a few hundred users, you're probably not touching that ceiling.
Where it bites is image-heavy sites with no proper CDN layer in front, or sites that get a sudden spike. I had a client, a musician, whose site was on my personal Vercel account during a transition period (wrong of me, I know, commercial use and all that). One of their tracks got picked up by a decent-sized playlist and the site got about 40,000 visits in 48 hours. Bandwidth usage shot up. Thankfully we were under the cap but it was close enough that I moved them off Hobby immediately.
If you're using Next.js Image Optimisation, be aware that optimised images served through Vercel count against your bandwidth. And the Hobby plan has a 1,000 source images cap on optimisation per month. That's a limit that almost nobody talks about and it will absolutely catch you if you're running a photography portfolio or any image-heavy content site.
The Commercial Use Thing Deserves More Attention
I want to come back to this because I think it's genuinely underappreciated in freelance circles.
Vercel's terms are clear. Hobby is for personal, non-commercial projects. The moment a client pays you to build something on Hobby, or the moment a project generates revenue (ads, subscriptions, sales), you're outside the terms. This isn't Vercel being aggressive, it's just how the plan is structured.
I've spoken to freelancers who deploy client work to Hobby to save money. Some have done it for years without consequence. But Vercel does enforce this. They can and do reach out to accounts running what appears to be commercial activity. Getting a site taken down because of a terms violation while a client's business depends on it is not a conversation you want to have.
The Pro plan is $20/month per member. For any actual client work, that cost should be baked into your project budget. It's not negotiable.
When the Single Concurrent Build Actually Hurts
One concurrent build doesn't matter for personal projects. It starts to matter when you're working on something collaboratively, even informally. Two people pushing commits in quick succession means the second build queues behind the first.
For solo work, this is a non-issue. But if you've got a contributor, a co-founder, a designer who can also commit, then you'll notice queue time. On Pro you get up to 12 concurrent builds. The jump is substantial.
Seahawk had a small internal tools project earlier this year where two of us were both iterating quickly on the same Next.js codebase during a sprint. We were on Pro for that one (for the commercial reason alone), and even so, overlapping builds occasionally created friction. On Hobby it would have been noticeably worse.
Hobby vs Pro: The Actual Decision Tree
Here's how I think about it, practically:
- Personal project, no revenue, solo dev, under 50 pages: Hobby is fine. Stay there.
- Side project getting real traffic (5,000+ monthly visits): Start watching your bandwidth and function execution hours. Hobby still works but you're closer to edges.
- Any client work at all: Pro. Full stop.
- API-heavy app where serverless functions regularly approach 5+ seconds: Pro. The 10-second timeout is not a comfortable ceiling at that point.
- Need multiple team members on the account: Pro. Hobby doesn't allow team accounts.
- Running e-commerce, subscriptions, or any transactional flow: Pro, and probably think carefully about whether Vercel's serverless model suits your architecture anyway.
For anything that outgrows even Pro, Vercel has an Enterprise tier with custom limits. I've only touched Enterprise twice, both times for clients with very specific SLA requirements. For most people reading this, the Pro/Hobby decision is the relevant one.
Alternatives Worth Knowing
If budget is genuinely the constraint, Vercel isn't your only option. Netlify's free tier has similar shape to Vercel Hobby with 300 build minutes per month (lower than Vercel, actually) but 100 GB bandwidth. For static sites with minimal serverless needs, Netlify is a reasonable comparison.
Cloudflare Pages is worth considering if your use case fits Workers. Unlimited bandwidth on the free tier, generous build allowances, and the Pages platform has matured significantly. I use it for static sites where I don't need Next.js-specific features.
Railway and Render are good for full-stack apps where you want more control than a serverless deployment model. Different tradeoffs, not direct Vercel competitors exactly, but worth having in your toolkit.
FAQ
Does the Vercel Hobby plan allow custom domains?
Yes. You can add custom domains on Hobby without paying anything. SSL is included automatically through Let's Encrypt. This is one of the genuinely great things about Hobby, there's no artificial limit on custom domains.
What happens when you hit the Hobby plan bandwidth limit?
Vercel doesn't immediately take your site down. They'll notify you and give you the option to upgrade. But you're in a grey zone once you're over and Vercel can throttle or suspend if you persistently exceed limits without upgrading.
Can I use Vercel Hobby for a client project if I'm not charging much?
The amount you charge is irrelevant to Vercel's terms. Commercial use means commercial use. If someone is paying you, it's commercial. Move to Pro.
Is the 10-second function timeout a hard limit or soft?
Hard. The function is killed at 10 seconds on Hobby. You'll see a 504 timeout error. There's no configuration available at the Hobby level to extend it.
Can two people collaborate on a Hobby project?
Not through a shared team account. Hobby is personal accounts only. You can add collaborators to a specific project (they can contribute code via Git), but the Vercel account and billing sits with one person. For actual team workflows, you need Pro.
---
The Hobby plan is genuinely good for what it says it is: hobby projects. The limits are reasonable for personal use. Where people get into trouble is treating it as a free Pro tier. It isn't. Know the caps, respect the commercial use terms, and when a project starts earning money or serving real users, pay the $20 a month. It's not worth the risk.