As enterprises transition from traditional data centres to "AI Factories"—high-performance hubs designed to convert raw data into autonomous business intelligence—the underlying architecture has become increasingly complex. Building these environments requires orchestrating diverse components, including high-density compute, specialised networking, and multi-vendor storage.

"Ensuring that security configurations and automations perform as expected across these layers is traditionally resource-intensive. The integration with NVIDIA DSX Air addresses this by providing a "digital twin" environment where organisations can perform large-scale cyber security validation without the risk or cost associated with physical migration errors," says Lionel Dartnall, Country Manager: SADC, Check Point Software Technologies.
Securing the New Threat Surface

The shift toward private Large Language Models (LLMs) and sovereign AI mandates has expanded the data centre threat surface. Traditional security measures are often insufficient against risks specific to AI workloads, such as:

    Model Integrity: Protecting against model theft, poisoning attacks, and prompt injections.

    Data Privacy: Preventing PII leakage via unprotected logs or illicit model responses.

    Lateral Movement: Blocking unauthorised "east-west" traffic within AI clusters.

    Infrastructure Vulnerabilities: Securing the AI supply chain and maintaining resilience across inference clusters and gateways.

Validation Through Digital Twins

Check Point's AI security stack operates within the DSX Air environment to provide full validation across the NVIDIA infrastructure, including Spectrum-X Ethernet switches, BlueField DPUs, and the NVIDIA AI software stack.

By running the AI Factory Firewall on simulated BlueField DPUs, organisations can test tenant isolation and granular security policies for both human users and AI agents. This ensures security is operational from "token number one," enabling a seamless transition to NVIDIA's high-performance hardware, including future-ready support for BlueField-4 DPUs expected in late 2026.
Significance for Data Centre Evolution

This integration marks a shift toward "Security by Design" for AI infrastructure. By enabling end-to-end testing of multi-vendor integrations—including Kubernetes and third-party storage—Check Point and NVIDIA allow enterprises to scale their AI capabilities while meeting mandatory government compliance and industry regulations.

"When building these proprietary AI factories, organisations often build from the ground up or undergo massive transformations of existing sites," said Lionel Dartnall, Country Manager: SADC at Check Point Software Technologies. "By using DSX Air, we enable them to validate every configuration and automation in advance, reducing the risks inherent in such significant architectural shifts."