Cloud company Akamai has launched two services for deploying AI models and running AI agents, as the cloud infrastructure provider expands its AI offerings for enterprises.
Akamai Inference provides managed access to open-weight and custom AI models through OpenAI-compatible APIs, while Akamai Agent Sandbox provides an isolated environment for executing code written by AI agents.
The services are part of Akamai’s strategy to offer infrastructure for building, running and securing AI applications on a single platform. The company said its cloud infrastructure spans GPU and CPU clusters in core regions and more than 4,350 edge locations.
“We have been implementing an AI infrastructure strategy based on the notion that AI inference will represent the majority of enterprise focus and investment. Our customers proved that thesis and pulled us further,” said Adam Karon, Chief Operating Officer and general manager of Akamai Cloud.
“They want to build, run, and protect AI on the same platform. Akamai Inference and Akamai Agent Sandbox add the inference and execution layers that turn our AI infrastructure into a full AI stack,” he added.
Akamai Inference gives enterprises 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 and Google Gemma 4, alongside embedding models for retrieval-augmented generation (RAG).
Customers can choose between serverless, pay-per-use inference and dedicated deployments. The service handles model execution, orchestration and authentication, while providing access to managed databases, object storage and connections to private and public cloud platforms.
The service uses OpenAI-compatible APIs, allowing developers to integrate it with existing applications without rewriting them. Akamai said requests run across its global network to reduce latency and support service availability.
Akamai Agent Sandbox provides an environment for executing code written by AI agents, separating it from the agent’s own runtime.
The service runs agents on Akamai Functions, while the code they generate executes in an isolated environment. It uses WebAssembly-based isolation to enforce limits on compute, memory, outbound network connections and system capabilities.
The sandbox is being released as an open-source component of the Spin framework. The company said it plans to add further capabilities to govern what AI 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.
“Some teams want to ship an AI feature this quarter on a managed platform. Others need dedicated GPUs to run frontier-class open models on their own terms,” he added.
Alongside the two services, Akamai has expanded its dedicated GPU fleet with NVIDIA Blackwell Ultra clusters and is preparing its data centres for NVIDIA’s Vera Rubin platform in 2027.
Its GPU infrastructure also includes NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs for low-latency inference. The dedicated clusters are intended for enterprises that want to build, fine-tune and host models on infrastructure they control.
Akamai said each eight-GPU Blackwell Ultra node provides 288 GB of HBM3e memory per GPU, or 2.3 TB across the node. The infrastructure is intended to support workloads involving reasoning, long-context processing and AI agents.
Dedicated deployments also give customers control over model versions and deprecation timelines, without rate limits or queueing associated with shared infrastructure.
The new services add managed inference and code execution to Akamai’s existing cloud infrastructure and security offerings, giving enterprises options to deploy models through managed services or run workloads on dedicated GPU infrastructure.
