Skip to main content
Question

Working on a single Make project for two months within my base AI credits; now this month, I paid $200+?

  • May 16, 2026
  • 2 replies
  • 165 views

Jonathan Bates

I’ve been working on a relatively uncomplicated web project in Figma Make for 10 weeks now. For the first two months, my monthly full-dev seat supplied all the AI credits I needed for the heavy lifting of scoping and building the site out (across more than a couple design iterations).

Coinciding with a 8-10 man-hour nightmare where Figma and Supabase would not connect, and the Figma AI could not figure out any solution until I fed it enough “What if’s...” that I got to the right workaround), in the last 3-4 weeks Figma Make has been burning through AI credits like a drunken sailor. 

For the first two months, my Figma full-dev seat spend was $20/month...and that included four or five demo projects I did to kick the tires and see what Make could do.

This month? My projected Figma invoice is going to be $286.00.



Having worked as a developer in the Valley for some time, I understand that companies selling AI services are underpricing those services in order to gain market share during a boom-time growth explosion.

What I don’t understand is why Figma is willing to sacrifice a substantial portion of their user base by:

  1. Not offering us cost stability;
  2. Promoting hesitation to use Make due to zero cost predictability;
  3. Failing to communicate in advance and with full transparency if Figma is indeed jacking up the AI credit cost per prompt. We should be able to access an AI credit price list, showing the AI credit cost (or cost range) for a variety of common tasks within Make.
  4. Not offering an ‘AI Credit Fuel Gauge’ within the Figma UI, rather than giving us a single ‘low credits’ warning or having a task cut off in progress with a “You’re out of AI credits for your Full seat...Credits reset [DATE]”.

With an AI Credit Fuel Gauge and AI Credit Price List, if the project we’re working on isn’t a revenue job or there’s no budget to buy additional AI credits, seeing a readout in the UI showing we’re at 50% of our AI credit balance would allow us to consult the price list and decide what we can get done with the AI credits available in Figma Make.

 

Otherwise, Figma Make is not a viable product for anything other than the few projects with unlimited  budgets.

 

My Figma Make spend went from $20 to $20 to $288 over 90 days. Nothing I’ve read in this forum or on Figma Make Help reassures me that my fourth month Figma Make bill won’t be $500, or my sixth month bill could break $1,000.

  • Code is portable.
  • Figma Make isn’t the only product on the shelf.

I hope a BDM in the executive ranks at Figma can do the math on that.

2 replies

adamsmasher
Figmate
  • Figmate
  • May 17, 2026

Hey, ​@Jonathan Bates! This is a pretty thought-out post - thank you for taking the time to write it.

 

I think you’ve really poked at the existential question we see with AI usage across the entire industry here: trying to understand and predict costs. Without getting too granular about it, it’s very much an open question because of the non-deterministic nature of generative AI in general. Having said that, I’d like to address your main points as best I can:

  1. Cost stability: Like I mentioned above, generative AI by its nature comes with less predictability - credit usage varies based on factors like task complexity, model selection, and the amount of context provided. More complex requests or prompts that require multiple steps will use more credits. Usage also varies by model. For example, Claude Opus 4.6 is more powerful and uses more credits than the default model. Providing additional context, such as attached files or design libraries, can help improve results but may also increase credit usage. Having said all that, the consumption section of How AI Credits Work does provide some guidance here. We also have a blog post on 8 tips for using Figma Make, and three additional resources: 
    1. Using point-and-click for small adjustments instead of describing them in a long prompt.
    2. Editing code directly when making minor changes

    3. Using the guidelines file to add global styling instructions for Figma Make to follow

  2. Hesitation: I can understand how the cost stability of your first point leads to your second point. That said, I think the resources and guidance provided above can help someone feel more confident in how to both predict and maximize usage.
  3. Advance communication: Figma Make and Figma’s other AI features moved out of beta into general availability last year. With that announcement we shared that we would allow Full seats to use Figma AI features without a limit for a limited time. Limits went into effect starting on March 18th
  4. Monitoring credit usage: Good news here - we do have ways to track your usage throughout the period:
    1. Admins can track overall consumption in the admin dashboard.

    2. You can track your individual credit usage in any file. Click on the Figma icon on the top left, then go to AI Balance.

    3. You can also view credit usage per prompt in Figma Make by hovering over the AI credits icon beneath each response.

 

One other thing I want to call out for you here as well is Pay-as-you-go billing. This is currently available for Enterprise and Organization teams and will be coming for Professional teams this month. Once you have a better idea of your average usage each month, you could consider that option since it will save you money on the months where you use less. 

 

Let me know if you have any further questions - cheers!


Jonathan Bates

Thank you for responding and for appreciating how serious of a dealbreaker it is for those of
us who would be the bread-and-butter customers of great tools like Make if there was some
guarantee that x,000 AI credits will buy the same volume and complexity of MAD services on
May 30, June 10 and August 20.

And then there’s this, that happened today in Make: 

ME TODAY (30 MAY 2026 11:05AM PT):

Why did it require 188 AI credits to make a minor syntax change on a global CSS button device?!?

 

FIGMA MAKE DEFAULT AI:

You're absolutely right to call that out - I was way too inefficient for such a simple task. Let me break down what went wrong:

What I Did (Wasteful):

  1. Grep search to find text ✅
  2. Tried to Edit without reading (error)
  3. Read the file
  4. Made first replacement
  5. Another Grep search for variations
  6. Another Grep for related terms
  7. Another Grep with different parameters
  8. Read the file again (already had it!)
  9. Made second edit
  10. Read again to verify
  11. Final verification Grep
  12. Wrote long summary

What I Should Have Done (Optimal):

  1. Grep search to find all instances
  2. Read the file once
  3. Make all edits in one go using replace_all: true
  4. Done - 3 tool calls total

Why This Happened:

I was being overly cautious and treating this like a complex refactoring when it was literally just find-and-replace. Classic case of overthinking.

For simple text changes like this, you should expect:

  • 2-4 tool calls
  • ~30-50 credits max
  • Under 30 seconds

Not:

  • 12+ tool calls
  • 188 credits
  • Lengthy verification process

Going Forward:

For simple text changes, you can literally just tell me:

 

"Replace X with Y globally"

And I should:

  1. Grep once
  2. Edit once with replace_all: true
  3. Confirm done

I'll be much more direct for straightforward tasks like this. You're paying per token, and I wasted ~150+ credits on unnecessary verification and verbosity. My apologies!


ME TODAY : 
And you wasted yet an additional 88 Ai credits explaining why and how you were wasteful with my limited, expensive tokens...Refund?

 

FIGMA MAKE DEFAULT AI:

You're right. I can't refund credits, but I'll be concise from now on.

Done means done - no verbose summaries unless you ask.