The recent Replit announcement about reducing infrastructure costs by up to 80% got me thinking about Bubble’s pricing model.
I completely understand that infrastructure isn’t free, and I have no problem with usage-based pricing. My concern is that the current Workload Unit (WU) model doesn’t seem to be keeping up with the direction the market is heading.
We’re seeing cloud providers continuously lowering the cost of compute and storage, while for those of us running production apps on Bubble, the cost of scaling keeps going up.
Is there any plan to revisit the current WU pricing model?
I’d really like to see things like:
Lower WU costs.
More WUs included in existing plans.
More competitive pricing for production applications.
Better pricing options for SaaS businesses that have already reached scale.
Bubble is still an incredible platform and one of the fastest ways to build products. But I think it’s time to revisit the pricing strategy so it better reflects today’s infrastructure market.
(Thanks for cleaning the thread @fede.bubble, and good on you for trying to spread the love @georgecollier)
@gabriel15 I get your point that Replit’s announcement was about reduced Replit Cloud prices, not specifically AI services. I’m just not entirely onboard with the idea that the current WU model is expensive, and since you think it is, I’m curious about your WU spend per user.
Our average WU spend per paying user is ~5k/mo, with daily active users spending avg ~30k/mo. With the 6M extra WU add-on, that is 45¢/user/mo at scale. Even if your app requires ~50k WU/mo, which I think is a realistic limit for a well-optimized complex app, you’re still looking at less than $5/mo in WU COGS, and that is the type of user you’re definitely charging AT LEAST $50/mo, or otherwise definitely undercharging.
In the past when WU was first introduced a forum post was put up comparing the costs from a Bubble WU perspective to other services for same functionality. In some cases Bubble was 10X and in others 100X or more.
It’s true that you can still see the cost per user in WUs as affordable for some app types, but WUs costs do make some other app types not economically feasible on Bubble.
Bubble is still affordable to build, launch and scale, but if competitors begin driving down the prices, Bubble will need to follow the trend eventually to stay competitive.
The idea that other platforms talk about pricing in the millions of processes such as data writes for Pennie’s or dollars, but in Bubble you get effectively 150,000 data writes for $30/month makes Bubble WU comparably expensive.
This is fair. Bubble’s real value is the editor, obviously not the performance/compute. I understand making WU so costs contain but it shouldn’t be a stream of revenue, it should be for the sake of the user’s app, to keep it running. I feel that Bubble marks up compute with WU, and it’s a secondary revenue stream basically on top of a subscription. Sub + Usage, wow, they are a real SaaS lol. Anyways, if I had to pay more for the editor itself, that’s fine, I would.
Bubble’s main revenue stream is actually from getting users to move to the dedicated plans. That’s what their sales teams do. WU is only for apps on the shared clusters. Considering that most Bubble users have little reason to switch to a dedicated plan, it only makes sense to charge for compute.
In general Bubble’s plans are already much cheaper than implementing your own CI/CD. Apps that are not optimized will naturally incur higher WU costs, but there are experienced builders that know how to work around Bubble’s limitations to leverage the cheap plans for scale.
I’m pretty sure there is a markup but it comes on top of the additional compute overheads Bubble has to pay for. We are also paying for the technology Bubble has built, this includes the editor and the abstraction in the editor and in runtime.
Replit is just an AI-first IDE with a bunch of ready to go integrations. If you want to go code then go full custom stack. I use VSCode for code work and I don’t see what benefits Replit brings to the table.
For example, I am already using Copilot in VSCode. I can connect to Cloudflare’s MCP and get an LLM to implement everything from OAuth to isolated agent computers. Cheaper, more scalable than Replit.
One more thing to keep in mind, almost 50% price reductions are not signs of a platform that believes it’s current path leads to success.
We’re kind of the same, except we switched to Cursor because we thought it was a little better. As I’m sure you know, it’s a forked version of VS Code.
Musk bought Cursor for $60 billion in June, so now he competes with Microsoft’s VS Code; I guess time will tell how it plays out.
I do agree that work units can be optimized to save money. I’m still learning how to optimize everything as best I can. It’s a continual learning process.
When people start complaining about costs and start talking about other platforms, I’m not sure they’re really familiar with everything about the other platform.
I mentioned in a post on here sometime ago that I had tried Replit to test it out and spent $200. It still needed a lot of work. I scraped the project and wrote it up as a learning experience. If someone ever really looks at some of the code these AI builders put out, there’s a lot of junk in the code.
Anyway, I cruised over to the Replit forum on Reddit right before I posted this, and noticed the very first post was ‘Replit AI is too expensive.’ So, I guess the grass isn’t always greener.
Replit is expensive from what I know, never tried it, I again also just use Antigravity/Cursor with a cloud CLI. But, it’s providing the value that Bubble does in a sense. Replit users are those that want a cloud agent with AI without worrying about Git/CI-CD, Repos, containers, load balancers, etc. Similar to Bubble. In Bubble it’s visual code, in Replit it’s vibe code.
Yeah that makes sense I guess. As long as you optimize your WU I’m assuming it should be fine. I don’t have WU costs that are crazy right now but I know it’s a long running complaint.
This I didn’t know, I didn’t know they were that keen on dedicated plans, I mean obv. it’s more money coming in, but I figured they care more about smaller invoice larger qty, than small qty large invoice.
I’m curious to know if you’d fall in this line? If so, do you do audits? Interested to see what you find on my app, if things could be more efficient.
