Akamai Builds a Full AI Stack Around Inference and Agents

Akamai is expanding its AI infrastructure. The company has launched Akamai Inference and Akamai Agent Sandbox, two services designed to help enterprises deploy AI models and run AI-generated code on one cloud platform.

Akamai Inference provides managed access to open-weight and custom AI models through OpenAI-compatible APIs, while Agent Sandbox creates an isolated environment for code written by AI agents. Together, the services add inference and execution layers to Akamai’s existing cloud infrastructure and security offerings.

That is the pitch, at least: build, run, tune, ground, and secure AI applications without stitching together a collection of unrelated services. Cloud platforms have discovered that “one platform” sounds especially attractive after customers spend six months connecting five of them.

Managed models across core and edge infrastructure

Akamai Inference offers access to 12 open-weight models ranging from 8 billion to 2.8 trillion parameters. The initial catalogue includes NVIDIA Nemotron 3 Ultra, OpenAI gpt-oss-120b, Google Gemma 4, and embedding models for retrieval-augmented generation, or RAG.

Customers can select serverless, pay-per-use inference or dedicated deployments. Akamai handles model execution, orchestration, and authentication, while its cloud provides managed databases, object storage, and connections to private and public cloud platforms.

Because the service uses OpenAI-compatible APIs, existing applications can move without rewrites. Requests run across Akamai’s global network, which the company says reduces latency and supports service availability and predictable performance.

That network spans GPU and CPU clusters in core regions and more than 4,350 edge locations. Akamai also offers managed Kubernetes, NVIDIA-accelerated infrastructure, dedicated CPU infrastructure, and Akamai Functions for serverless execution across edge and core environments.

Adam Karon, chief operating officer and general manager of Akamai Cloud, said the company has been building its strategy around the expectation that AI inference will account for most enterprise focus and investment. “They want to build, run, and protect AI on the same platform,” Karon said.

A sandbox for code agents and dedicated GPUs

Akamai Agent Sandbox addresses a different problem: what happens when an AI agent writes code that must execute. Agents run on Akamai Functions, but the generated code runs in a separate isolated environment rather than inside the agent’s own runtime.

The sandbox uses WebAssembly-based isolation to enforce limits on compute, memory, outbound network connections, and system capabilities. Akamai is releasing it as an open-source component of the Spin framework and plans to add controls governing what agents can access and which systems can interact with them.

“Every team running agents needs a boundary around the code those agents write,” said Jon Alexander, senior vice president of cloud technology at Akamai. His distinction is practical: some teams want to ship an AI feature this quarter on a managed platform, while others need dedicated GPUs to run frontier-class open models on their own terms.

For the second group, Akamai has expanded its dedicated GPU fleet with NVIDIA Blackwell Ultra clusters. Its infrastructure also includes NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs for low-latency inference, and the company is preparing its data centres for NVIDIA’s Vera Rubin platform in 2027.

Each eight-GPU Blackwell Ultra node provides 288 GB of HBM3e memory per GPU, or 2.3 TB across the node. Akamai targets that capacity at reasoning, long-context processing, test-time scaling, and AI agent workloads—tasks that have a habit of turning memory requirements into someone else’s budget problem.

Dedicated clusters give enterprises control over model versions and deprecation timelines without the rate limits or shared-infrastructure queueing associated with shared deployments. They are intended for customers that want to build, fine-tune, and host models on infrastructure they control.

Akamai Inference and Akamai Agent Sandbox are available in beta, with serverless deployment options included in the Inference beta. The company says the services build on Akamai Cloud, which has secured multiyear contracts collectively worth more than $14 billion.

The broader strategy is clear: combine GPUs, CPUs, storage, networking, edge functions, managed inference, agent execution, and security on one interconnected platform. Whether enterprises want managed access or dedicated control, Akamai is positioning its infrastructure as the place where AI applications move from model experiments into production.

Stay connected via Google News

Follow us for the latest AI News