Akamai Technologies Inc. has expanded its AI Cloud platform by introducing two new components: Akamai Inference and Akamai Agent Sandbox. These additions aim to provide enterprise customers with managed environments for deploying artificial intelligence workloads and agents across cloud data centers and edge locations.
Akamai Inference is a managed service offering access to open-weight and custom AI models via OpenAI-compatible APIs. At launch, users can choose from 12 models, including Nvidia Nemotron 3 Ultra, OpenAI gpt-oss-120b, and Google Gemma 4, with parameter counts ranging from 8 billion to 2.8 trillion.
The Akamai Agent Sandbox provides an isolated execution space for autonomous AI agents using Akamai Functions and the open-source Spin framework. This environment allows enterprises to enforce granular control over compute, memory, system APIs, and network access to ensure secure operations.
The company is also expanding its GPU hardware offerings within the AI Cloud. It now provides access to Nvidia Blackwell Ultra clusters, specifically RTX PRO 6000 Blackwell Server Edition GPUs, where each eight-GPU node delivers up to 2.3 terabytes of HBM3e memory.
Akamai plans to further increase its hardware capacity by adding Nvidia’s next-generation Vera Rubin GPU platform. This new hardware is scheduled to become available to customers in the new year.
