We expanded NVIDIA RTX PRO™ 6000 Blackwell GPU capacity across our global cloud infrastructure and introduced 8-GPU configurations for larger AI inference, accelerated graphics, and simulation workloads.
Valkey Managed Database adds a high-performance, in-memory option for applications that depend on fast access to data, including AI applications that need caching, retrieval, and operational memory.
Updates to LKE Enterprise, LKE, and App Platform strengthen the foundation for running AI workloads at scale, with improvements to networking, application management, and day-to-day operations.
Reserved public IP addresses, placement group aware maintenance, private image sharing, single sign-on (SSO), and monitoring improvements give teams more control over how infrastructure is deployed and managed.
If you've been following our quarterly updates, you've seen how we're continuing to expand Akamai Cloud, both in the capabilities we're bringing to market and the infrastructure we're building to support them.
In Q1, we expanded our AI infrastructure with NVIDIA RTX PRO™ 6000 Blackwell GPUs and introduced Akamai Functions.
In Q2, we focused heavily on strengthening the foundation around that compute, including strengthening networking, identity, observability, Kubernetes, storage, and other capabilities you need to run applications in production.
In this post, we'll explore how we're building on that progress by expanding AI compute capacity, introducing new data services, and making it easier for you to deploy, scale, and operate applications on Akamai Cloud.
AI workloads are getting larger, and the infrastructure requirements are becoming more varied.
This quarter, we expanded our NVIDIA RTX PRO™ 6000 Blackwell GPU infrastructure with new 8-GPU configurations. The additional GPU and VRAM capacity gives you more room to run larger models; supports longer context windows; and handles demanding language, multimodal, graphics, and simulation workloads. We’ve also recently made available Akamai B300 GPU, powered by NVIDIA HGX B300 with Blackwell Ultra GPUs.
At the same time, we continued expanding GPU capacity across our global cloud infrastructure. As more organizations move AI into production, they need access to accelerated compute in the markets where they operate. Our global footprint puts us in a unique position to support that expansion, giving you greater flexibility in where and how you deploy AI workloads.
The direction is straightforward: Give customers more choices for how much accelerated compute they need and give them more places to run it.
As applications become more distributed and increasingly data-intensive, fast access to data becomes critical. Whether you're building real-time applications, managing session state, or supporting AI workloads that depend on context and memory, the performance of the data layer directly affects the experience you can deliver.
In Q3 2026, we introduced Valkey Managed Database, a managed, high-performance in-memory data service built on the open source Valkey ecosystem. Valkey supports caching, session storage, and other real-time data requirements, while also enabling emerging AI use cases, such as caching prompts and responses, storing application state, and supporting retrieval-augmented generation (RAG).
As AI applications move into production, the relationship between compute and data becomes increasingly important. Fast inference depends on having timely access to the context and information that applications need. Valkey is another step in expanding the data services available on Akamai Cloud to support these evolving workloads.
Kubernetes continues to be an important part of how you deploy and operate workloads on Akamai Cloud, and we're evolving LKE Enterprise to support the growing demands of production AI.
In Q3 2026, we continued to improve LKE networking and cluster operations, including completing the transition of existing LKE clusters from WireGuard to Konnectivity for communication between the Kubernetes control plane and worker nodes. This brings existing clusters onto the same networking architecture already used by new clusters.
These updates are part of a broader evolution of LKE Enterprise for AI workloads. LKE Enterprise can support clusters with up to 5,000 GPUs, giving you a Kubernetes foundation for running AI workloads at significant scale. We're also investing in the capabilities around that compute, including GPU-aware scheduling, high-performance storage, and an open AI stack built around technologies such as KServe, Kubeflow, Istio, and Knative.
As AI infrastructure and tooling continue to evolve, Kubernetes provides an open foundation for deploying and managing these workloads. Rather than tying your applications to a proprietary AI stack, you can build on familiar open source technologies and adopt new tools and frameworks as your requirements change.
Building on that foundation, App Platform v6.0 and v6.1 continued to simplify how you build, deploy, and operate applications on Kubernetes. Q3’s releases included improvements to secrets management, GitOps, authorization, multi-tenant observability, and core platform components.
As applications scale, the infrastructure that is distributing traffic to them needs to scale too. In Q3 2026, Premium NodeBalancers became generally available, running on dedicated infrastructure to provide more predictable performance and greater scale. Premium NodeBalancers run on dedicated infrastructure and support up to 100,000 concurrent connections and 2,000 back-end nodes, along with TCP, HTTP/HTTPS, and UDP load balancing.
We also expanded how Akamai NodeBalancers connect to back-end infrastructure. Both Basic and Premium NodeBalancers can now connect to back ends using VPC or public IPv6, giving teams more flexibility in how they design and isolate application infrastructure.
Premium NodeBalancers are also automatically assigned to LKE Enterprise clusters, strengthening the networking layer underneath larger Kubernetes and AI workloads.
The evolution of NodeBalancers is another part of making Akamai Cloud ready for applications when scale, predictable performance, and network architecture matter in production.
Production infrastructure changes constantly. Instances are rebuilt. Applications move. Maintenance happens. Teams automate deployments and replace resources. The surrounding infrastructure shouldn’t have to be rebuilt every time that happens.
Reserved public IPv4 addresses, introduced in Q3 2026, separate the public IP from the individual VM or NodeBalancer using it. The address stays associated with the account until it is explicitly released, even if the underlying resource is deleted or rebuilt.
That means you can replace or migrate infrastructure within a region without also changing the public-facing IP, updating DNS records, rewriting firewall rules, or modifying allowlists and automation.
It's a relatively simple capability, but it removes an operational dependency that becomes increasingly important as infrastructure becomes more automated.
The same principle applies to maintenance. Placement groups already give customers a way to spread instances across physical hosts for high availability. In Q3 2026, we made scheduled maintenance aware of those placement decisions.
For workloads using anti-affinity placement groups, Akamai Cloud can limit simultaneous scheduled maintenance across instances that perform similar functions.
Availability isn’t only about how infrastructure is deployed. It also depends on what happens to that infrastructure over its lifetime. Making the platform aware of those relationships helps preserve the architecture you’ve designed when maintenance occurs.
As cloud environments grow, teams also need repeatable ways to distribute infrastructure across users and accounts. Private Image Sharing, now in beta, allows teams to securely share custom images with specific users rather than manually duplicating or uploading those images for each account.
For organizations that use golden images to standardize operating systems, software, configuration, or security requirements, Private Image Sharing provides a simpler way to maintain consistency while still controlling who can access those images.
These updates, combined with capabilities such as SSO and the role-based access control (RBAC) improvements we've delivered this year, are allowing us to continue to reduce the amount of manual work required to manage infrastructure across larger organizations.
As applications become more complex, having visibility into your infrastructure becomes increasingly important. In Q3 2026, we continued to expand Akamai Cloud's observability capabilities with improvements across metrics, dashboards, and alerts.
Updates to observability include Valkey metrics, service-specific dashboard presets, email-based alert recipients, PDF dashboard exports, improved alert details, and additional ways to group metric data.
These improvements make it easier to understand what's happening across your infrastructure. As applications become more distributed across compute, Kubernetes, databases, networking, and AI infrastructure, having a consistent view of performance and operations becomes increasingly important.
The work we've delivered over the first three quarters of 2026 is making Akamai Cloud increasingly ready for the demands of enterprise workloads.
We've expanded accelerated compute infrastructure for AI, added a serverless developer layer with Akamai Functions, and continued to strengthen the networking, Kubernetes, identity, observability, storage, and data services that support applications in production.
Looking ahead, we're seeing a significant shift take shape across the industry. As AI moves into production, inference is becoming an increasingly important part of the market. Models are growing larger and more capable, while the rise of AI agents is driving more frequent and complex interactions with those models. Together, these trends are changing the scale and nature of inference workloads and creating new demands for accelerated compute, real-time data, secure execution, and application infrastructure.
These shifts are shaping where we're taking Akamai Cloud next. We're continuing to expand the scale and capabilities of our AI infrastructure while investing in new ways to make it easier for you to build, deploy, and operate inference and agentic applications.
We look forward to sharing more about what's coming in the months ahead.
© 2026 Akamai Technologies
