On 5 November 2026, AWS announced the general availability of DynamoDB bulk put items, a fully managed feature that writes data from Amazon S3 into your existing DynamoDB tables. If you have ever built a Glue job, an EMR cluster or a rate-limited Lambda writer just to push a nightly file from S3 into a live table, this launch is aimed squarely at you.
This guide explains what DynamoDB bulk put items does, how it differs from what was available before, how to run your first import safely, and what it costs.
What Changed With DynamoDB Bulk Put Items?
DynamoDB already had Import from S3, but that feature only creates a new table. As the DynamoDB Developer Guide puts it, Import from S3 creates a new table from data in S3, while bulk put items writes into a table that already exists. Until now, loading files into a table that was serving traffic meant building a custom pipeline and carefully managing capacity so bulk writes did not compete with your application.
According to the AWS Database Blog launch post, bulk put items works like this:
- It runs as a managed, asynchronous job. You point it at an S3 location with files in line-delimited DynamoDB JSON or Amazon Ion, optionally compressed with gzip or zstd.
- Writes use dedicated system capacity, so they do not consume your table’s read or write capacity and do not count toward your read and write CloudWatch metrics. Your existing alarms and auto scaling rules keep working unchanged.
- Each record can carry its own condition expression, using the same grammar as
PutItem. - Records that are not written are listed in a report delivered to an S3 bucket you choose.
- The input files can come from one account while the target table lives in another.
Imported items behave like ordinary writes. Global and local secondary indexes are updated automatically, and items appear in DynamoDB Streams and Kinesis Data Streams as if they were application writes, so downstream consumers and zero-ETL integrations receive them too.
Where Bulk Put Items Fits
AWS says the most common customer pattern is refreshing machine learning insights. A personalization system produces fresh inference data every day, often hundreds of gigabytes, and it has to land in the serving table without slowing down the application. Other good fits named in the launch post include:
- Enrichment: updating millions of product SKUs after a price change.
- Backfills: populating a new attribute that a feature needs, using archived data.
- Migrations: loading exported data, with a version check so stale items never overwrite newer application writes.
It is not a replacement for PutItem or BatchWriteItem. Each operation spends its first few minutes provisioning, so if you need a handful of items written in seconds, the regular APIs are still the right tool.
How to Run Your First Bulk Put Items Operation
1. Prepare the input files
Write one record per line. Each record holds an Item and, optionally, a ConditionExpression. When a record has no condition, the item is written unconditionally and overwrites any existing item with the same key. Use attribute_not_exists(pk) for insert-only writes, or compare a version attribute so only newer copies are written:
{"Item": {"pk": {"S": "SKU#1001"}, "price": {"N": "499"}}, "ConditionExpression": "attribute_not_exists(pk)"}
{"Item": {"pk": {"S": "SKU#1002"}, "price": {"N": "799"}, "Version": {"N": "12"}}, "ConditionExpression": "#v < :v", "ExpressionAttributeNames": {"#v": "Version"}, "ExpressionAttributeValues": {":v": {"N": "12"}}}
An item skipped by its condition is not counted as a failure. The service parallelizes across input objects, so split large datasets into several similarly sized S3 objects instead of one huge file, and keep partition keys high-cardinality and evenly spread.
2. Set up permissions
The operation runs with the caller’s permissions. The identity needs the account-scoped dynamodb:StartBulkOperation action, the table-scoped dynamodb:BulkPutItems action, read access to the input bucket, write access to the report bucket, and access to any AWS KMS keys involved.
3. Choose a tier and start the job
You can start from the console, the AWS CLI or the SDKs. The launch post shows this CLI example:
aws dynamodb start-bulk-put-items \
--table-name arn:aws:dynamodb:us-east-1:123456789012:table/MusicCollection \
--input-s3-configuration '{"S3Bucket":"my-import-bucket","S3KeyPrefix":"daily/","InputFormat":"DYNAMODB_JSON"}' \
--output-s3-configuration '{"S3Bucket":"my-report-bucket","S3KeyPrefix":"reports/","OutputType":["ERROR","CONDITION_CHECK_FAILURE"]}' \
--performance-tier ACCELERATED
The performance tiers documentation explains the choice. Standard is the default and suits work with flexible timing, such as background enrichment. Accelerated targets higher throughput for migrations, recovery scenarios and jobs with a fixed window. AWS estimates that, on a well-scaled table with no secondary indexes and evenly distributed keys, Accelerated can write 10 million 1 KB items in under 10 minutes, 100 million in under an hour, and 1 billion in 2–4 hours. Accelerated is available in a list of Regions that includes Asia Pacific (Mumbai) and Asia Pacific (Hyderabad), which is useful for Indian teams.
4. Monitor the operation, not the table
Because bulk writes do not show up in your table’s consumed-capacity metrics, your table alarms will not tell you if an import went wrong. Track the job with describe-bulk-operation instead. It ends in one of four states: COMPLETED, COMPLETED_WITH_ERRORS, FAILED or CANCELLED. Cancelling does not roll back items already written, and list-bulk-operations keeps 90 days of history. Every API call is logged in AWS CloudTrail.
5. Reconcile with the report
Each report record shows the source S3 object, the record’s position in it and the reason it was not written. For condition check failures, it shows both the incoming item and the item already in the table. Treat COMPLETED_WITH_ERRORS as normal: fix the failed records and re-import only those.
Costs and Limits to Plan For
Bulk put items uses byte-based billing: you pay per GB written, with a 1 KB minimum per item, charged separately for the table and for each index the write reaches. In US East (N. Virginia), the launch post lists Standard at $0.10 per GB and Accelerated at $0.20 per GB, so importing 100 million 1 KB items costs about $10 on Standard or $20 on Accelerated. Items rejected by validation are not billed, but items that fail a condition check are, and report objects are billed at normal S3 rates.
Key limits at launch:
- One bulk operation per table at a time, and none while an index is being created.
- Up to five Standard and five Accelerated operations in flight per account.
- Up to 100 GB per operation on Standard and 1 TB on Accelerated.
- Global tables are not supported yet.
- It writes whole items only. For attribute-level changes, use
UpdateItem. - For imports under 1 GB, AWS recommends Standard, because Accelerated adds little benefit at that size.
If the table feeds Streams consumers, size them for your normal write rate plus the import, or cap the job with the MaxWriteRequestUnits parameter. Set that cap carefully, because a very low value can push the operation past its time limit and cause it to fail. If you are replacing an existing Glue or EMR pipeline, our AWS data analytics solutions team can help you plan the switch, and our AWS migration services cover version-safe data moves into DynamoDB.
Frequently Asked Questions
Does DynamoDB bulk put items slow down my application?
No. AWS says the writes run on dedicated system capacity and consume none of your table’s read or write capacity, so application traffic is not affected.
How is bulk put items different from Import from S3?
Import from S3 creates a new table. Bulk put items writes into a table that already exists and is serving traffic.
Can I re-run the same file safely?
Yes, if you add conditions such as attribute_not_exists or a version comparison. A client token stops a retried start call from launching a duplicate, but a deliberate re-run is a new operation, so only your conditions protect newer data.
Which tier should I choose?
AWS’s advice is simple: choose Standard unless the operation has a deadline, and choose Accelerated when it does.
Retire Your Custom Ingestion Pipeline
Shanav Technologies helps startups and growing businesses in India build simpler, cheaper data pipelines on AWS. If you want to replace custom S3-to-DynamoDB loaders with a managed, auditable process, talk to our team and we will help you roll out DynamoDB bulk put items safely.