Amazon SQS marks 20 years with new AI & scaling tools

Amazon Web Services has marked the 20th anniversary of Amazon Simple Queue Service, one of the first three AWS products launched to customers.

Amazon SQS was created to let software components exchange messages without requiring direct, synchronous connections between services. That basic design remains the main reason customers use the product, even as it has gained new controls, security settings, and support for more complex workloads over the past two decades.

The service was introduced alongside Amazon EC2 and Amazon S3 in the early years of AWS. It addressed a common problem in distributed systems: if components were too tightly linked, one slow or unavailable service could trigger failures across a broader application.

Message queuing offered a different model. One part of an application could place a message in a queue and continue its work, while another part could process that message later when ready. The approach helped isolate faults and absorb spikes in demand.

Recent changes

Over the past five years, AWS has focused on increasing throughput, changing encryption defaults, and adding tools for message recovery and queue administration.

One of the most significant changes involved FIFO queues, which preserve message order. In 2021, AWS introduced a higher-throughput mode that supported 3,000 transactions per second per API action. It then raised that limit several times, reaching 70,000 transactions per second per API action in selected regions by late 2023.

AWS also changed the security baseline for new queues. In 2021, it introduced server-side encryption using SQS-managed keys, removing the need for customers to manage their own encryption keys for that option. A year later, it made that setting the default for all newly created queues.

AWS also expanded recovery options for messages in dead-letter queues, which hold messages that could not be processed successfully. It first added the ability to move messages back to the source queue from the SQS console, then extended that function to its software development kits and command-line tools through dedicated application programming interfaces. It later added the same redrive support for FIFO queues.

Access control has been another area of development. In 2022, AWS added attribute-based access control for SQS, allowing permissions to be defined through queue tags rather than fixed policies tied to individual resources. The model is intended to simplify administration as deployments scale.

Broader use

Several changes have focused on efficiency and integration. Support for the JSON protocol in the AWS SDK was added in 2023, reducing end-to-end message processing latency by up to 23% for a 5 KB payload while also lowering client-side CPU and memory use, according to AWS.

That same year, AWS added direct integration from the SQS console to Amazon EventBridge Pipes, giving users a way to route queue messages to other AWS services without writing custom integration code.

Support for larger messages has also expanded. In 2024, AWS released an extended client library for Python that lets developers send messages of up to 2 GB by storing the main payload in Amazon S3 and passing a reference through SQS. It also increased the in-flight message limit for FIFO queues from 20,000 to 120,000, enabling more concurrent processing before consumers hit the previous ceiling.

In 2025, AWS introduced what it called fair queues for multi-tenant standard queues. The feature is designed to limit the effect of a noisy neighbour so one tenant does not delay delivery for others when users include a message group ID when sending messages. AWS also raised the maximum payload size for standard and FIFO queues from 256 KiB to 1 MiB, with parallel support added in AWS Lambda event source mapping.

AI workloads

While SQS's core role remains conventional application messaging, the same pattern is now being used in artificial intelligence systems, according to AWS. Customers are using queues to buffer requests to large language models, manage inference throughput, and coordinate exchanges between autonomous AI agents running as separate services.

That reflects a broader trend in cloud infrastructure, where long-established middleware services are being adapted for newer AI use cases rather than replaced outright. In this case, the queue continues to act as a buffer and control point between producers and consumers, even when those consumers are models or software agents rather than traditional application services.

SQS has been part of AWS since the platform's earliest commercial phase, and its continued expansion shows how foundational infrastructure services can remain relevant by adapting to new operational patterns while keeping their original purpose intact. Customers use SQS to decouple services, buffer bursts of traffic, and build systems that remain resilient when individual components fail.