Akamai Cloud Is Built for What Cloud Has Become (Updated May 2026)

AI is moving into production systems. Inference now sits directly in the path of the application, where latency, consistency, and cost all matter simultaneously.

In Q1 2026, we expanded our GPU offering with the launch of NVIDIA RTX PRO™ 6000 Blackwell GPUs by building on our existing RTX 4000 Ada Generation instances.

Customers are running more demanding workloads in production across computer vision, speech, multimodal applications, and real-time inference. In a market where high-end GPU capacity is constrained, access to that infrastructure matters.

The GPU is only part of the story; where that compute runs matters just as much.

When inference can run closer to where requests originate and where data already exists, system behavior changes in measurable ways:

• Latency drops.

• Cross-region traffic is reduced.

• Response times become more consistent.

• Use cases such as real-time assistants, personalization, and decision systems move from prototypes into production systems.

GPUs handle inference, but everything around those GPUs still depends on CPU performance.

APIs, orchestration, data processing, and integration layers all rely on consistency. When that layer is unpredictable, the entire system is affected.

With our G8 dedicated plans, built on 5th Gen AMD EPYC™ processors and backed by dedicated hardware, we deliver consistent, non-oversubscribed performance. No noisy neighbors. And no guessing how workloads will behave under pressure.

As these systems come together, a more balanced model takes shape: Run inference on distributed GPUs where it makes sense; run the rest of the application on CPU infrastructure that delivers consistent results.

Up to this point, the platform has focused on infrastructure; that is, the compute, storage, networking, and the services needed to run distributed systems.

With Akamai Functions, our edge native serverless platform built on WebAssembly, we’re adding a developer layer on top.

Modern applications don’t run in one place. They require decisions about where logic executes and how requests are handled. Akamai Functions gives developers a way to run code across both edge and core without managing the underlying infrastructure.

Akamai EdgeWorkers already provides low-latency execution at the edge. Akamai Functions extends this with broader language support (e.g., SDKs for languages like Rust, Go, JavaScript, and Python) and more control over where workloads run.

In practice, this means handling requests closer to users, keeping heavier logic near data and compute, and using GPUs more efficiently by offloading everything around inference. 

Akamai Functions is part of how we keep the experience simple without limiting what can be built.

In addition to expanding into GPUs and Akamai Functions, we’ve recently focused on improving access control, visibility, and operational consistency.

We introduced role-based access control (RBAC) to move from coarse permissions to a more granular, standardized model. This makes it easier to manage access across teams, handle onboarding and offboarding, and meet compliance requirements without adding operational overhead.

We also expanded observability. Real-time metrics are now available for Databases, Object Storage, and NodeBalancers. Audit logs are available across core regions, with alerts for Object Storage and Databases, and broader coverage expanding over time.

For teams running across edge services, delivery, and security, TrafficPeak extends that visibility across the full stack.

Kubernetes continues to be a primary way that teams run and operate applications on Akamai Cloud.

LKE, Akamai’s managed Kubernetes engine, continues to stay current with ongoing version updates and infrastructure improvements. Recent releases include upgrades across multiple Kubernetes versions, infrastructure improvements, and foundational networking updates such as Konnectivity.

At the application layer, updates like App Platform v4.13.0 focus on making the platform more reliable and easier to operate. Moving to a modern operator-based deployment model, removing deprecated components, and improving upgrade stability and Argo CD operations reduces the operational overhead for teams running applications in production.

We expanded our object storage footprint with new E3 endpoints in Chicago, Frankfurt, Los Angeles, and Tokyo. E3 endpoints are designed for high-throughput access to object storage, so placing them in more regions gives teams more options to keep data closer to where it’s being read and written, reducing latency and avoiding unnecessary cross-region movement.

We also added support for metrics and audit logs, so teams can see how data is being accessed and track changes across environments without stitching together external tooling.

That same focus on global data access and predictable cost models is starting to show up externally as well, including in the Forrester Object Storage Solutions Landscape.

At the same time, we continued to evolve the underlying data and networking layer. Valkey Managed Database is now in beta for teams looking for Redis-compatible performance and APIs without the licensing constraints, making it easier to run high-throughput, in-memory workloads without adding cost or complexity.

Managed Databases also now support dual-stack virtual private clouds (VPCs) in limited availability, allowing workloads to run across both IPv4 and IPv6 environments without reworking how services connect as networks evolve.

AI inference is changing how applications are built and run. More of the work is happening in real time, closer to users, and across more distributed systems.

That shift introduces real constraints. Latency budgets tighten, workloads become more fragmented, and systems have to coordinate across compute, data, and execution in ways that centralized models were not designed for.

At the same time, reliability, observability, and core infrastructure performance remain baseline requirements for running these systems in production.

We are building the platform to support that model. 

That’s the update for this quarter. We’ll keep building and share our progress in the next quarter.

In the meantime, you can try Akamai Cloud for yourself.