Nested unstable_cache results are not cached.

[khobiziilyes]

Link to the code that reproduces this issue

https://github.com/khobiziilyes/unstable-cache-reprod

To Reproduce

1. Start the app in preview mode (npm run preview).
2. Open the main page (http://localhost:3000/)
3. Visit Brand 1 page (http://localhost:3000/1)
4. Visit Brand 2 page (http://localhost:3000/2)

Current vs. Expected behavior

I've tried to simplify the flow of the issue as much as possible.

The project consists of 6 main commits. The first 2 are unimportant (t3app bootstrapping). The 3rd commit introduces the minimal and most basic structure of the app, which is:

The app has 3 main "api" functions (apis.ts)

• getBrands: returns an array of brandId.
• getBrandDetails: receives the id of the brand, returns the name and description.
• getSuggestedBrands: For simplicity, it returns all brands except the one you supply. In a real case scenario, it would calculate the similarity to the other brands using the Levenshtein algo. Notice that THIS IS the critical function, since it implements another function, getBrands, inside it.

All of those have a console.log to simplify debugging.

The main page of the app (/) displays a list of links to all the brands available, which are retrieved using getBrands().
Each Brand page (/[brandId]) displays the details of the current brand, and also displays other suggested brands to visit which is a Suspended Server component that uses getSuggestedBrands.

If we run our app (npm run preview) and visit the main page, we'll notice that getBrands is not called (The whole page is static on build, the code is not running at all).

Now, we want to view a single brand page. If we visit the brands 1, 2, and 3 respectively, we'll see those logs:

And this is perfectly normal since we are not doing ISR.
Notice that each getSuggestedBrands calls getBrands and uses that value.

Now, since those "api" calls are expensive, we have decided to add a cache layer to those functions, and since they are not "fetch" calls. We'll use unstable_cache instead (not React.cache because we want it to be shared among all requests). And that's what the 4th commit is for, we wrapped the getBrands and getBrandDetails in unstable_cache. We will rebuild the app and visit the brands pages again, we will notice those logs:

We can notice that getBrands is not called anymore, because it is already called at build time and cached. And now the cached value is used in all calls, including the ones that getSuggestedBrands invokes.

If we refresh the page /3, we will even notice that getBrandDetails(3) is not called anymore, since the first visit to the url invoked and cached the value. Super!

However, we have one more function left that is not cached, and in real life, such function will consume a lot of processing power, and we'd love to cache the final result of it. getSuggestedBrands, this function has only 1 parameter, currentBrandId, which it uses to find more interesting brands similar to that one.

Commit 6: Introduces the issue.

After we wrap the function (which inherently uses another cached function getBrands), visit the pages, we'll get those logs:

getBrands is called over and over ... If we try to use this function on its own, we'll get the cached value, but if we use it inside getSuggestedBrands, the cached value is not retrieved.
getSuggestedBrands is cached as expected.

Provide environment information

This project is configured to use npm because /Users/ilyesk/Developer/unstable-cache-reprod/package.json has a "packageManager" field

Operating System:
  Platform: darwin
  Arch: arm64
  Version: Darwin Kernel Version 23.6.0: Thu Mar  6 22:08:50 PST 2025; root:xnu-10063.141.1.704.6~1/RELEASE_ARM64_T8112
  Available memory (MB): 16384
  Available CPU cores: 8
Binaries:
  Node: 22.15.0
  npm: 11.4.1
  Yarn: N/A
  pnpm: N/A
Relevant Packages:
  next: 15.4.0-canary.82 // Latest available version is detected (15.4.0-canary.82).
  eslint-config-next: 15.3.3
  react: 19.1.0
  react-dom: 19.1.0
  typescript: 5.8.3
Next.js Config:
  output: N/A

Which area(s) are affected? (Select all that apply)

Use Cache

Which stage(s) are affected? (Select all that apply)

next start (local), Vercel (Deployed), next dev (local)

Additional context

Before each preview, I run rm -rf .next/ to make sure no cached values will mess with the reproduction.

I haven't dug deep into Next.js code, but I'd like to help resolve this issue if possible.

[github-actions]

We could not detect a valid reproduction link. Make sure to follow the bug report template carefully.

Why was this issue closed?

To be able to investigate, we need access to a reproduction to identify what triggered the issue. We need a link to a public GitHub repository (template for App Router, template for Pages Router), but you can also use these templates: CodeSandbox: App Router or CodeSandbox: Pages Router.

The bug template that you filled out has a section called "Link to the code that reproduces this issue", which is where you should provide the link to the reproduction.

• If you did not provide a link or the link you provided is not valid, we will close the issue.
• If you provide a link to a private repository, we will close the issue.
• If you provide a link to a repository but not in the correct section, we will close the issue.

What should I do?

Depending on the reason the issue was closed, you can do the following:

• If you did not provide a link, please open a new issue with a link to a reproduction.
• If you provided a link to a private repository, please open a new issue with a link to a public repository.
• If you provided a link to a repository but not in the correct section, please open a new issue with a link to a reproduction in the correct section.

In general, assume that we should not go through a lengthy onboarding process at your company code only to be able to verify an issue.

My repository is private and cannot make it public

In most cases, a private repo will not be a sufficient minimal reproduction, as this codebase might contain a lot of unrelated parts that would make our investigation take longer. Please do not make it public. Instead, create a new repository using the templates above, adding the relevant code to reproduce the issue. Common things to look out for:

• Remove any code that is not related to the issue. (pages, API routes, components, etc.)
• Remove any dependencies that are not related to the issue.
• Remove any third-party service that would require us to sign up for an account to reproduce the issue.
• Remove any environment variables that are not related to the issue.
• Remove private packages that we do not have access to.
• If the issue is not related to a monorepo specifically, try to reproduce the issue without a complex monorepo setup

I did not open this issue, but it is relevant to me, what can I do to help?

Anyone experiencing the same issue is welcome to provide a minimal reproduction following the above steps by opening a new issue.

I think my reproduction is good enough, why aren't you looking into it quickly?

We look into every Next.js issue and constantly monitor open issues for new comments.

However, sometimes we might miss one or two due to the popularity/high traffic of the repository. We apologize, and kindly ask you to refrain from tagging core maintainers, as that will usually not result in increased priority.

Upvoting issues to show your interest will help us prioritize and address them as quickly as possible. That said, every issue is important to us, and if an issue gets closed by accident, we encourage you to open a new one linking to the old issue and we will look into it.

Useful Resources

• How to Contribute to Open Source (Next.js)
• How to create a Minimal, Complete, and Verifiable example
• Reporting a Next.js bug
• Next.js Triaging issues