Lambda Code Storage Default Limit Raised to 300GB: Exceeding the Old 75GB Limit Without a Quota Increase Request
This page has been translated by machine translation. View original
Introduction
On July 15, 2026, the default limit for AWS Lambda managed code storage was raised.
| Item | Old default | New default |
|---|---|---|
| Managed code storage limit | 75 GB | 300 GB |
Until now, the initial value was 75GB per Region, and a quota increase request was required to go beyond that. With this change, you can use up to 300GB without submitting a request.
What I Tested
Test Environment
| Item | Value |
|---|---|
| Region | ap-northeast-1 |
| Quota increase request | None |
| Existing Lambda functions | 17 |
| Dummy zip size | 209,715,368 bytes (about 200MB) |
| Number of layers created | 390 |
| Layer compatible runtime | python3.12 |
Checking with get-account-settings
First, I checked the current limit using get-account-settings.
aws lambda get-account-settings --region ap-northeast-1
{
"AccountLimit": {
"TotalCodeSize": 322122547200,
"CodeSizeUnzipped": 262144000,
"CodeSizeZipped": 52428800,
"ConcurrentExecutions": 1000,
"UnreservedConcurrentExecutions": 1000
},
"AccountUsage": {
"TotalCodeSize": 68507563,
"FunctionCount": 17
}
}
AccountLimit.TotalCodeSize is 322,122,547,200 bytes (300GB). It has been raised from the previous default of 80,530,636,800 bytes (75GB). The size limits for individual functions, such as CodeSizeUnzipped (250MB) and CodeSizeZipped (50MB), have not changed.
The usage at the start of the test (AccountUsage.TotalCodeSize) was 68,507,563 bytes (about 65MB).
Demonstrating Usage Beyond 75GB
I created a dummy zip file of about 200MB. I generated 200MB of random data from /dev/urandom and archived it with zip -0 (uncompressed store).
dd if=/dev/urandom of=dummy.bin bs=1M count=200
zip -0 dummy.zip dummy.bin
Upload the created zip to S3.
aws s3 cp dummy.zip s3://<BUCKET_NAME>/<OBJECT_KEY>
Create a layer by referencing the zip on S3. The following is an example of a single run.
aws lambda publish-layer-version \
--layer-name dummy-layer-001 \
--content S3Bucket=<BUCKET_NAME>,S3Key=<OBJECT_KEY> \
--compatible-runtimes python3.12 \
--region ap-northeast-1
I created 390 layers in this manner. In practice, I ran them in bulk with a shell loop.
for i in $(seq -w 1 390); do
aws lambda publish-layer-version \
--layer-name "dummy-layer-$i" \
--content S3Bucket=<BUCKET_NAME>,S3Key=<OBJECT_KEY> \
--compatible-runtimes python3.12 \
--region ap-northeast-1 &
done
wait
The time required for layer creation alone was about 7 minutes.
| Timing | TotalCodeSize | Notes |
|---|---|---|
| Before creating layers | 68,507,563 bytes (≒65MB) | 17 existing functions |
| After creating 390 layers | 81,857,501,083 bytes (≒76.24GB) | Exceeded the old 75GB limit, no errors |
| After cleanup | 68,507,563 bytes (≒65MB) | Returned to the original usage |
Usage exceeded the old default limit of 75GB (80,530,636,800 bytes) by about 1.24GB, but CodeStorageExceededException did not occur. All layers were created successfully. After the test, when I deleted all the dummy layers, the usage returned to its original value.
Note that if you upload directly with publish-layer-version using --zip-file, you will hit the CodeSizeZipped (50MB) limit. Via S3 (--content S3Bucket=...,S3Key=...), I was able to create layers without problems even with a 200MB zip. However, even via S3, the 250MB limit after extraction (CodeSizeUnzipped) still applies.
Summary
The default limit for AWS Lambda managed code storage has been raised from 75GB to 300GB.
In this test, I created 390 dummy layers in an account without a quota increase request and confirmed that usage could be increased to about 76.24GB, exceeding the equivalent of the old 75GB limit. Even when exceeding 75GB, CodeStorageExceededException did not occur, and layer creation completed successfully.
Previously, a quota increase request was needed before exceeding 75GB, but from now on you can use up to 300GB without a request. In environments that deploy many functions and layers, this reduces the hassle of submitting requests and lets you manage code with more headroom. Note that this 300GB quota is currently listed as non-adjustable.
If 300GB is still not enough, you can also use REFERENCE mode, which is not counted toward code storage. For details, see the following article.
