2026-10-11 16:37 UTC

AWS claims its limited-rollout project spend limits cap monthly pre-tax bills by stopping project resources, providing an infrastructure-level runaway-cost control for experiments and interruption-tolerant agent workloads.

state: watchingheat: lowuncertainty: mediumconvergesscott: mediumcloud-infrastructure inference-economics agent-governanceAWS

What is this?

AWS documents a project-level spend-limit feature in AWS Settings, describing it as setting the most a developer will pay per month and pausing the project when its limit is reached. The official search snippet is truncated at its description of resource stopping; it does not establish limited availability, pre-tax accounting, optional earlier termination controls, or the exact scope of the cap. Separate AWS Budgets documentation describes alerts and periodically updated cost tracking, not this project's pause mechanism; applying the new feature to interruption-tolerant agent workloads is the case's proposed use, not a documented AWS claim in these snippets.

Why it matters to Scott

AWS’s documented project pause at a spending limit converges with Scott’s Autonomy Budget, offering a potential infrastructure-level enforcement primitive to evaluate for his AWS Marketplace Knowledge Appliance rather than another cost alert. The radar tracks related whole-run budget enforcement in TokenOps, not this AWS development; applicability to his deployment, exact cap coverage, limited availability and pre-tax treatment remain unestablished.
ip:concept.autonomy-budgetip:framework.architecture-not-vibesdev:project.applianceradar:tokenops-whole-run-budgetradar:concept.agent-governance
queries asked of Scott's wikis
  • agent runaway spend hard budget enforcement
  • infrastructure guardrails versus agent self-governance
  • AWS experiment sandbox cost isolation
  • agent workload interruption recovery checkpointing
  • inference economics budget caps versus cost alerts

Measured heat

now 0 pts/hpeak 2 pts/hcomments 0/hpeers p14momentum: steady2 platformsage 566h
points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion

How the heat travelled

09-18 02:30 (minted)⭐ origin echo-reconstructedDocuments limited-availability project-level spend limits that stop resources at the limit, with optional earlier controls that can terminat
AWS on blog (echo) · attributed from hn.story.49749357 · published time unknown
—
09-18 01:59first on hacker news · published · lag ?AWS now supports spend limits
xyos
—
09-18 01:59amplified on hacker news 👑hn.story.49749357
xyos
peak 22 · 4 comments · 96% of case engagement
10-05 12:43amplified on hacker newshn.story.49964116
speckx
peak 1 · 0 comments · 4% of case engagement
09-18 02:20our radar first saw it · lag ?discovery anchor: hn.story.49749357—
pace: p54 vs 1032 stories at the 336h mark (now 566h old) — ahead of autobot-persistent-chatgpt-harness (1.1x), behind breadcrumb-flight-recorder-agent-memory (0.9x)

Evidence (3) — ⭐ canonical anchor

sourceobjectauthorscorecomments
🟧 hnAWS now supports spend limits
Retrieved article excerpt

Open article · Retrieved 2026-09-18T02:21:59.523835+00:00

[View a markdown version of this page](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.md)

Create a spend limit in AWS Settings - AWS Account Management

[Documentation](https://docs.aws.amazon.com/index.html)[AWS Account Management](https://docs.aws.amazon.com/organizations/index.html)[User Guide](https://docs.aws.amazon.com/accounts/latest/reference/accounts-welcome.html)

[What is a spend limit?](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.html#what-is-spend-limit)[Set your spend limit](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.html#set-your-spend-limit)[Notifications](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.html#spend-limit-notifications)[Customize your spend limit settings with optional early cost controls](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.html#spend-limit-early-controls)[What happens if you reach your limit](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.html#spend-limit-reached)[Other important things to note](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.html#spend-limit-other-notes)[Create a spend limit in AWS Settings](https://docs.aws.amazon.com/accounts/latest/reference/create-spend-limit.html#create-spend-limit-procedure)

# Create a spend limit in AWS Settings

###### Warning

We're currently releasing our new experience to a limited number of customers. You might not be able to access this experience yet.

## What is a spend limit?

Spend limits enable developers to use AWS with confidence that they can operate within
a predictable and controlled budget. Use a spend limit to set the most you'll pay per month
for a project. Spend limits are at the project level. You can have spend limits on some
of your projects and have other projects without spend limits.

If your usage in a project reaches its limit, AWS pauses that project which stops
its resources so your costs stay within it. Spend limits are designed for experimentation,
learning, and sandbox workloads. They can be used for production environments when it is
acceptable to have a brief pause of your resources in cases when your applications incur
unexpected costs.

Spend limits are visible to anyone with project access, but can only be managed by
project owners. You must be on a Paid Plan to create a spend limit. For more information,
see [Upgrade your account in AWS Settings](https://docs.aws.amazon.com/accounts/latest/reference/upgrade-account.html).

Spend limits are a type of [AWS
Budgets](https://docs.aws.amazon.com/cost-management/latest/userguide/budgets-managing-costs.html).

## Set your spend limit

Your spend limit is the ceiling on your project's pre-tax costs, not a fee. When you
set a spend limit, the minimum allowed value is the greater of $20 or a conservative
estimate of your likely spend.

**Why a minimum exists:** Hitting a spend limit is a
disruptive experience. The minimum exists to make it unlikely that normal variations in your
usage will lead to hitting the spend limit.

**How the conservative estimate is calculated:** The
estimate is based on your activity month to date, but it's not a strict linear
extrapolation. AWS looks at your spend to date and the resources you currently have
running. AWS also uses your activity from the prior month — if that was high, it becomes
the minimum.

**If you have many resources running:** The conservative
estimate of your likely spend will be high. If you want to set a lower spend limit, you
need to stop those resources first.

**If you are reactivating a paused project:** The
conservative estimate of your likely spend is your spend to date during the month plus any
usage that was not charged.

###### Note

Spend limits exclude credits and apply to pre-tax charges.

## Notifications

You'll receive notifications when actual costs reach 50%, 75%, and 90% of your limit,
or when you are on track to reach or exceed your spend limit in the next 10 days.

## Customize your spend limit settings with optional early cost controls

Your spend limit keeps your bill from growing past the amount you set. If your usage
reaches it, AWS pauses your resources to protect you. These optional controls act earlier
and more gently, so you stay well below the limit.

### Stop new resource creation

About 7 days before you would reach your limit, new resources stop launching. This is
achieved through application of a service control policy that is managed by AWS. The
service control policy depends on the service. To view all service control policies that
might be applied to your project, see [Service control policies for spend limits](https://docs.aws.amazon.com/accounts/latest/reference/scps-and-rcps-for-projects.html#scps-for-spend-limits).

All your resources that are already running stay running. In most cases, nothing you
have built is affected. However, in some cases this could lead to a disruptive experience.
For example, if you have enabled auto-scaling, and if conditions are met to execute this
action, new instances won't launch.

### Pause idle resources

About 5 days before you would reach your limit, idle resources are paused. Your active
apps keep running. Idle resources are paused based on recommendations by AWS Compute
Optimizer. For more information, see [View
idle recommendations](https://docs.aws.amazon.com/compute-optimizer/latest/ug/view-idle-recommendations.html).

You can also view your idle resources recommendations by logging into your project
and viewing savings opportunities. For more information, see [Cost
Optimization Hub](https://docs.aws.amazon.com/cost-management/latest/userguide/cost-optimization-hub.html). At this time, as part of your spend limits experience we will
automatically pause EC2, RDS and SageMaker endpoints.

**Why we pause an EC2 instance:** We pause EC2 instances
if the peak CPU utilization is below 5% and your network I/O is less than 5 MB per day
over the last 14 days.

If you have a G or P instance type, there's a different idle criterion. G or P instance
types will be paused if the following is true over the last 14 days:

- GPU isn't actively working for more than 99% of the lookback period
- GPU encoder isn't used for 99% or more of the instance's runtime
- GPU memory usage at instance level is less than 5%
- CPU maximum utilization is less than 5%
- Network utilization is less than 5 MB/day

**Why we pause an RDS database:** We pause RDS for MySQL
and RDS for PostgreSQL if the database instance is not a read replica and has the following
over the past 14 days:

- No database connections
- Low CPU usage
- Low read/write activity

**Why we pause an idle SageMaker endpoint:** We pause
SageMaker endpoints if the endpoint had zero invocations in the past 14 days.

### Pause top cost drivers

About 4 days before you would reach your limit, your highest cost incurring active
resources are paused. This can be a disruptive experience. At this time, we select the
resources from these five services: EC2, RDS, Lambda, Bedrock, and SageMaker. This opt-in
control might prevent you from incurring costs if you have a runaway Lambda or unexpected
Bedrock spike.

**How we pause a top cost driver EC2 instance:** When an
EC2 instance is identified as a top cost driver, AWS terminates the instance to eliminate
compute costs. Before termination, AWS creates a snapshot of each attached EBS volume to
preserve your data. Any associated Elastic IP addresses are also released.

After the instance is terminated:

- All compute and EBS volume costs are eliminated
- Your data is preserved in snapshots, which incur a small storage cost
- Elastic IP charges stop

To restore your workload, you can launch a new instance from the saved snapshots. If
you no longer need the data, you can delete the snapshots to stop snapshot storage
charges.

**How we pause a top cost driver RDS database:** When an
RDS database instance is identified as a top cost driver, AWS stops the instance.
Stopping the instance eliminates compute charges while preserving your data.

While the instance is stopped:

- No compute charges are incurred
- Storage and provisioned IOPS charges continue
- Automated backups continue
- RDS automatically restarts the instance after 7 days — AWS will stop it
  again if the spend limit is still at risk

To restore your database, start the instance from AWS Settings or the RDS console.
Your data and configuration remain intact.

**How we pause a top cost driver Lambda function:** When a
Lambda function is identified as a top cost driver, AWS disables its event source
mappings and triggers to prevent further invocations. If the function has provisioned
concurrency configured, AWS also removes it to eliminate those charges.

After the function is paused:

- No invocation charges are incurred
- The function code, configuration, and all associated resources remain intact
- No data is lost

To restore your function, re-enable the event source mappings and triggers. If you had
provisioned concurrency, you'll need to reconfigure it.

**How we pause a top cost driver Bedrock model:** When an
Amazon Bedrock provisioned model is identified as a top cost driver, AWS deletes the
provisioned throughput. There is no pause option for provisioned throughput — deletion is
the only way to stop charges.

After the provisioned throughput is deleted:

- Provisioned throughput charges stop immediately
- If you have a custom fine-tuned model, the model weights are deleted
- Your original training data in Amazon S3 is not affected

To restore a custom model, you'll need to retrain it from your S3 training data. For
provisioned throughput on foundation models, you can create a new provisioned throughput
allocation.

## What happens if you reach your limit

AWS pauses your project and stops all resources. Your data is preserved.

To reactivate your project, increase your spend limit from AWS Settings. After reactivation, some
resources may need to be manually restarted.

###### Important

If you take no action within 90 days of your project being paused, AWS permanently deletes
your project data.

## Other important things to note

- If you use Amazon Route 53, you can only purchase one domain every 24 hours.
  You must ensure that domain costs are within spend limits, otherwise, your domain purchase
  won't go through. In such cases, you must increase spend limits before making a
  purchase.
- You can apply spend limits to up to 10 projects.

## Create a spend limit in AWS Settings

###### To create a spend limit in AWS Settings

1. Open AWS Settings at <https://settings.aws.com>.
2. In the main navigation pane, choose **Billing**.
3. In **Cost by project**, for **Spend limit**,
   choose **Set limit** for a project.
4. You can either choose the recommended spend limit or a custom limit.
5. If you choose a custom limit, enter the custom amount.
6. Choose **Set spend limit**.
7. To customize your spend limit with early cost controls, you can select
   **Stop new resource launches**, **Pause idle
   resources**, or **Pause top cost drivers**.
8. Choose **Save early cost controls**.

The spend limit will take effect immediately.

[Document Conventions](https://docs.aws.amazon.com/general/latest/gr/docconventions.html)

Upgrade your account

Update a spend limit

Did this page help you? - Yes

Thanks for letting us know we're doing a good job!

If you've got a moment, please tell us what we did right so we can do more of it.

Did this page help you? - No

Thanks for letting us know this page needs work. We're sorry we let you down.

If you've got a moment, please tell us how we can make the documentation better.
xyos224
🟧 echo.blog ⭐Documents limited-availability project-level spend limits that stop resources at the limit, with optional earlier controls that can terminatAWS——
🟧 hnOne runaway AI agent racked up a $50k cloud billspeckx10

Interpretation history

Decision trace