Akamai Technologies, a cybersecurity and cloud company, said on Wednesday it is working with Nvidia to introduce a security architecture for AI factories.

The companies will integrate Akamai Guardicore Segmentation into the Nvidia Vera BlueField-4 STX storage architecture, which runs on the Nvidia DOCA software platform.

The partnership centres on embedding a zero trust architecture as a default layer in AI factories to help enterprises protect data, context memory and autonomous agents.

Akamai explained that the move would allow AI factory operators to apply workload-based segmentation, monitor agent behaviour and block threats at the infrastructure layer. It added that security can run at the same speed as accelerated computing without burdening the GPU, CPU or storage cycles that AI workloads rely on.

Ofer Wolf (오퍼 울프), senior vice president of Akamai's enterprise security unit, said, "As frontier LLM-based attacks increase the speed and scale of cyber threats, AI factories are core assets that must be designed on the premise of blocking threat spread." He said, "In environments where every clock cycle matters, existing host-based security tools are like speed bumps on a race track. By applying workload-based segmentation to Nvidia Vera BlueField-4 STX and DOCA, we are helping implement zero trust at the speed of AI workloads themselves and block threats before they spread across the entire high-performance environment."