Akamai adds managed services to AI stack

Akamai has introduced two foundational models for its AI stack designed to help enterprise build and tune open-weight models, including their own models.

Akamai Inference and Akamai Agent Sandbox allow customers to ground open weight models in their own data, run agents inside enforced boundaries, and secure their AI-related applications, agents and infrastructure.

Collectively, the capabilities add managed services on top of the infrastructure used by Akamai’s customers to serve, run and secure AI applications and agents using both CPUs and GPUs. Akamai’s global network fabric also provides cloud storage, networking and edge functions for the purpose.

With the additions to the ecosystem, Akamai Cloud can provide users with access to their data, routing and governance and a hard boundary around what code developed by AI agents can reach, according to Akamai Cloud COO and GM Adam Karon.

“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,” he said. “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, and our security portfolio protects every layer of it. Akamai is uniquely positioned to bring our scale and distribution to address scaling real-time inference and agentic workloads.”

Akamai Inference meanwhile gives enterprises managed access to open weight and custom models through OpenAI-compatible APIs. Customers can choose serverless, pay-per-use inference or dedicated deployments. At launch, the catalogue includes 12 open weight models ranging from 8 billion to 2.8 trillion parameters.

The company also revealed it has expanded its dedicated GPU fleet to include NVIDIA Blackwell Ultra GPU clusters, and is preparing to support NVIDIA Vera Rubin next year. Each eight‍-‍GPU Blackwell Ultra node carries 288 GB of HBM3e memory per GPU for a total bandwidth of 2.3 TB, enough to hold frontier-class models, Karon said.

Image credit: iStock.com/Just_Super