Akamai expands its AI Cloud platform with dedicated inference and agent execution environments

Akamai Technologies Inc. is expanding its AI Cloud infrastructure platform with two new core components called Akamai Inference and Akamai Agent Sandbox.

They provide their enterprise customers with a highly customizable managed environment from which they can deploy and run artificial intelligence workloads and agents on any infrastructure, from the cloud data centers to the network edge.

The company says it is addressing growing enterprise demand for a more flexible yet reliable delivery infrastructure and a way to integrate AI with proprietary data sources and third-party data sources with strict governance protocols. As autonomous AI agents take on more responsibility for software coding, companies will also need stricter boundaries to protect these systems.

The Akamai AI Cloud is designed to do all that and more, giving customers everything they need to get their AI workloads and agents running: computing power, cloud storage, networking and security. It also offers a wide range of GPUs and other AI accelerators available at more than 4,350 edge locations. The Akamai Cloud Platform is at the heart of the company’s AI strategy, which calls for AI inference to occur closer to end users and the data they need to minimize latency.

Given its ambitious goals, the launch of Akamai Inference is a big deal. According to Akamai, it is a managed service offering that gives customers access to dozens of open and custom AI models through OpenAI-compatible application programming interfaces. The main benefit of Akamai Inference is that teams can deploy new AI applications and agents at the edge without having to rewrite code.

At launch, companies can choose from 12 open-weight models with 8 billion to 2.8 trillion parameters, including Nvidia Corp.’s Nemotron 3 Ultra, OpenAI Group PBC’s gpt-oss-120b, and Google LLC’s Gemma 4. There is also a selection of special “embedding” models for pipelines to produce advanced data retrievals. These are intended to connect the open weight models with Akamai’s managed database services to ensure their answers are based on enterprise data.

The Akamai Agent Sandbox, on the other hand, is all about secure execution. It is an isolated execution space for running AI agents that work autonomously on behalf of human users. Within the sandbox, AI agents run on Akamai Functions, with their code contained in isolated WebAssembly environments managed by the open source Spin framework. These sandboxes allow organizations to enforce granular control over everything from agent processing power and memory to their system APIs and outbound network access, ensuring they never overstep their bounds.

Adam Karon, chief operating officer and general manager of Akamai Cloud, said the company expects inference to become the primary focus for the vast majority of enterprises in the future. “Our customers have already proven this thesis,” he said. “You want to build, run and protect AI on the same platform. Akamai Inference and Akamai Sandbox add the inference and execution layers that transform our AI infrastructure into a complete AI stack, and our security portfolio protects every layer of it.”

The company also wants to give customers more choice in the hardware they use for their AI workloads, and has also announced that it is expanding its fleet of GPU platforms within the Akamai AI Cloud. It now offers access to Nvidia Blackwell Ultra clusters, specifically RTX PRO 6000 Blackwell Server Edition GPUs. Each eight-GPU Blackwell Ultra node offers up to 2.3 terabytes of total HBM3e memory, enabling support for long-context reasoning, Akamai said.

Akamai also plans to further expand its fleet with Nvidia’s next-generation Vera Rubin GPU platform, which should be available in the new year, it said.

Image: SiliconANGLE/Dreamina