AWS April 2026: Infrastructure Gets Simpler, AI Gets Governed

AWS has been busy with multicloud connectivity, S3 as a file system, and AI tooling reaching production readiness. Here is what matters for your organisation.

AWS has had a busy couple of weeks. Between new connectivity services, storage innovations, and a wave of AI tooling reaching general availability, there is a lot for organisations to absorb. Here is what caught our attention and what it means for businesses running workloads on AWS.

Multicloud connectivity, managed properly

AWS Interconnect is now generally available. The service provides managed private connectivity between Amazon VPCs and VPCs on other cloud providers. Alongside the multicloud capability, AWS has introduced Interconnect last mile, which simplifies high-speed private connections from your premises to AWS.

This matters because most organisations we work with are not single-cloud. They run workloads across providers, sometimes by design, sometimes by acquisition. Until now, stitching those environments together with private, reliable connectivity meant a patchwork of Direct Connect circuits, partner solutions, and manual configuration. AWS Interconnect turns that into a managed service.

For organisations already running hybrid or multicloud architectures, this is worth evaluating. Simpler connectivity reduces operational overhead and, importantly, reduces the surface area for misconfiguration.

S3 buckets as file systems

Amazon S3 Files lets you mount S3 buckets as high-performance file systems on AWS compute resources. Latencies sit around 1ms. The goal is to remove the long-standing trade-off between object storage benefits (durability, cost, scale) and the interactive file access that many applications and workflows require.

This is a meaningful change for teams running data pipelines, machine learning training jobs, or legacy applications that expect a POSIX file interface. Previously, you might have maintained both S3 and EFS, duplicating data and adding complexity. S3 Files could let you consolidate. We would recommend testing thoroughly before migrating production workloads, particularly around permission models and consistency behaviour, but the direction is clearly right.

AI tooling matures: agents, guardrails, and cost control

Several AI-related announcements landed in quick succession, and together they paint a picture of an ecosystem moving from experimentation to production.

Claude Mythos preview in Amazon Bedrock. Anthropic's latest model is now available for preview through Bedrock, giving organisations access without managing infrastructure directly.
AWS Agent Registry. A central place to catalogue and manage AI agents across your organisation. As agent-based architectures proliferate, discoverability and governance become essential.
AWS DevOps Agent and Security Agent reached general availability. These purpose-built agents handle operational and security tasks within AWS environments. GA status signals AWS considers them production-ready.
Bedrock Guardrails now supports cross-account safeguards. Organisational guardrails can be centrally enforced across multiple AWS accounts within an AWS Organisation. This is a significant step for enterprises that need consistent AI safety controls without relying on each team to configure them independently.

The cross-account guardrails feature deserves particular attention. One of the biggest risks we see with generative AI adoption is inconsistent governance. Team A configures content filters. Team B does not. A centralised, enforceable policy model addresses that gap directly.

AWS also flagged a theme we hear constantly in our own client conversations: cost visibility. As AI workloads move from proof-of-concept to production, finance and engineering teams need shared visibility into what these services actually cost. If you are scaling AI on AWS, getting your tagging strategy and cost allocation right now will save painful conversations later.

What organisations should consider

Review your connectivity architecture. If you operate across multiple clouds or have complex on-premises links, AWS Interconnect could simplify your setup and improve your security posture.
Evaluate S3 Files for storage consolidation. If you are running parallel object and file storage, there may be an opportunity to reduce cost and complexity.
Centralise your AI guardrails. If you are running Bedrock workloads across multiple accounts, enable organisational guardrails now rather than retrofitting them later.
Get ahead on AI cost management. Implement consistent tagging, set up AWS Cost Explorer dashboards for AI services, and establish accountability before spend surprises you.
Catalogue your agents. If your teams are building AI agents, the Agent Registry provides a governance layer. Use it early.

At Metaphor, we help organisations adopt AWS services in ways that are secure, well-governed, and cost-effective. If any of these announcements raise questions about your own environment, we are happy to talk them through.

Sources: AWS Interconnect is now generally available, AWS Weekly Roundup: April 13, 2026, Launching S3 Files, AWS Weekly Roundup: April 6, 2026, Amazon Bedrock Guardrails cross-account safeguards