Overview
Explore
Resources
Secure AI Infrastructure with Check Point and NVIDIA
Our strategic collaboration enables enterprises and neocloud providers to securely develop and deploy AI solutions across the AI factory and data center.
As organizations worldwide build AI factories, they face unprecedented challenges in protecting this infrastructure. Traditional security tools were never built to address threats like data exfiltration, data poisoning, denial of service, and model inversion at AI-scale. Check Point AI Cloud Protect integrates with NVIDIA BlueField™-3 DPUs to deliver industry-leading threat prevention while preserving peak AI performance, giving enterprises and service providers a secure environment to develop and deploy AI at scale.
- Robust Security Against AI Threats: tamper-proof code, network, and host-level security to prevent data poisoning, model theft, and exfiltration
- Validated for Everyday Enterprise Use: validated on NVIDIA RTX PRO™ Servers across the AI lifecycle: model development, agentic applications, and inference
- Future-Ready Infrastructure: security runs natively on NVIDIA BlueField DPUs today, with BlueField-4 performance gains on the roadmap
Securing AI Cloud Infrastructure with Check Point and NVIDIA
Check Point AI Cloud Protect integrates with NVIDIA BlueField DPUs to secure AI infrastructure at scale, providing industry-leading threat prevention while ensuring peak AI performance. The collaboration enables service providers and enterprises to develop and deploy AI models and applications in a secure environment.
- Validated on NVIDIA RTX PRO Servers
- AI Cloud Protect is open for enterprise deployment, validated on NVIDIA RTX PRO servers. This validation demonstrates enterprise-ready security for critical AI workloads including AI model development, agentic AI applications, and inference, built on NVIDIA BlueField technology. Organizations can deploy AI Cloud Protect on NVIDIA RTX PRO servers today.
- Ready for Next-Generation AI Infrastructure
- AI Cloud Protect runs on NVIDIA BlueField-3 today and is designed for the newly announced NVIDIA BlueField™-4, which delivers 6x the compute power and 800 Gb/s throughput. BlueField-4 also adds NVIDIA DOCA™ Flow and NVIDIA DOCA™ Vault for tenant segmentation and data access control. This next-generation platform supports 4x larger-scale AI factories and 2x performance for AI inference and storage processing, so security scales with AI infrastructure rather than becoming the bottleneck.
Key Capabilities of AI Factory Firewall
- Network-level protection: identity-based threat prevention with virtual patching and IPS, blocking unauthorized access to AI infrastructure before it reaches the workload
- Host-level protection: NVIDIA DOCA™ Argus uses direct memory access to monitor every running process on the AI node, detecting and stopping malicious processes, including compromised code hidden in publicly downloaded models
- Application-layer protection: CloudGuard WAF defends AI applications against prompt injection, jailbreaking, and LLM poisoning, plus runtime protection for RAG pipelines and MCP servers, covering LLM inputs, outputs, and data flows
- Powered by Check Point ThreatCloud AI: over 50 AI engines correlating threat intelligence from more than 150,000 connected networks and millions of endpoints, feeding the same threat intelligence that protects Check Point's broader portfolio into AI Cloud Protect
- Runs entirely on the NVIDIA BlueField DPU, isolated from the host CPU and GPU, so security adds zero CPU or GPU overhead and eliminates the performance tradeoff of traditional security tools
- NVIDIA DOCA Argus inspects and stops runtime attacks in real time at the DPU, without adding latency to the AI training or inference workload itself
- Validated on NVIDIA RTX PRO servers across the full AI lifecycle: model development, agentic AI applications, and production inference
- Built for NVIDIA BlueField-4 as well, which delivers 6x the compute power and 2x the network throughput of BlueField-3, so performance headroom grows with the hardware rather than the security layer becoming the bottleneck
- Unified management and automation for security orchestration across thousands of AI nodes, so policy scales with the AI factory instead of requiring node-by-node configuration
- Multi-tenant policy management applies distinct security policies to different tenants, AI agents, and users from a single console
- No manual agent installation on individual hosts, since protection runs on the DPU rather than on the AI server itself
- The same DOCA Argus-based protection extends from centralized AI factories to distributed edge and small-language-model (SLM) deployments under one policy, with no separate architecture required
- Kubernetes-native orchestration: NVIDIA's DOCA Platform Framework (DPF) manages the AIFF container on each BlueField-3 DPU. A Data Center Provider owns the infrastructure and provisions a dedicated management domain per tenant (Group Policy Controller, CloudGuard Controller), while each End-Customer installs their own AIFF container and manages their own security policy, isolating tenants within shared infrastructure. Zero Touch deployment auto-connects new AIFF gateways and installs policy with no manual per-device setup.
- Beyond Kubernetes, CloudGuard Controller (the same component used for workload discovery) also integrates with AWS, Azure, GCP, Cisco ACI, Cisco ISE, OpenStack, and VMware vCenter/NSX as data center sources. Kubernetes is the only one of these confirmed step-by-step in the AI Factory Firewall documentation; the others are general CloudGuard Controller capabilities not yet confirmed as wired into AI Cloud Protect specifically, and are a less natural fit given its bare-metal, BlueField DPU-based architecture.
- AI Cloud Protect runs inside the NVIDIA DSX Air simulation environment, operating on simulated BlueField DPUs alongside NVIDIA Spectrum™-X Ethernet switches, NVIDIA ConnectX™ SuperNICs, and the broader NVIDIA AI software stack
- Enterprises can validate security policies, including multi-tenant policy enforcement for both AI agents and human users, on a full digital-twin replica of their AI factory before it goes live
- Runs pre-deployment cyber security checks against that replica to catch vulnerabilities before production, ensuring security is enforced starting at token number one
- Demonstrated by Check Point and NVIDIA at GTC 2026, with tenant isolation for the AI data center highlighted as a core capability
- Built for NVIDIA BlueField-3 today; already engineered for NVIDIA BlueField-4, expected in early availability during 2026 as part of NVIDIA's Vera Rubin platform generation.
- BlueField-4 delivers 6x the compute power and 800 Gb/s of throughput versus BlueField-3, built around an NVIDIA Grace™ CPU and paired with NVIDIA® ConnectX®-9 SuperNIC™ for ultra-low-latency networking. Vera CPU and paired with NVIDIA® ConnectX®-9 SuperNIC™ for ultra-low-latency networking. NVIDIA Vera is the CPU for agents, delivering 80% faster agentic CPU performance for longer-thinking agents, faster-learning RL systems, and data-driven AI factories.
- NVIDIA states BlueField-4 supports AI factories up to 4x larger in scale than BlueField-3 allows, so as customers scale to gigascale AI factories, security scales with the hardware instead of becoming the constraint.
- DOCA Flow adds runtime workload visibility, tracking traffic and workload behavior across the AI factory in real time.
- DOCA Vault adds secure tenant segmentation, data access controls, and infrastructure-level protections purpose-built for AI factory environments, extending Check Point's multi-tenant policy model down to the DPU.
Why WWT & Check Point
Professional & Technical Services Opportunity
DPU-level deployment, architecture review, and security-posture assessment create differentiated services engagements for every AI factory build.
Open Platform & NVIDIA Ecosystem Alignment
AI Cloud Protect rides the same NVIDIA BlueField DPU / NVIDIA DGX ™ / RTX PRO server motion WWT already sells, a natural security attach rather than a competing line item.
High-Growth Market Aligned to NVIDIA Infrastructure Sales
Every DGX cluster and neocloud deployment is a security opportunity; AI Cloud Protect sits parallel to the NVIDIA GPU server sales motion already active at WWT.
Explore Check Point Content
Hands-On Lab Workshop: Check Point SD-WAN Lab
Join us for a hands-on look at Check Point's SD-WAN solution, which centralizes routing decisions across multiple WAN links and automatically steers traffic to the best-performing path in real time. In this workshop, you'll see how Check Point builds encrypted tunnels that connect branches, data centers, and cloud networks under a single, centrally managed security policy — improving application performance, increasing resiliency through link diversity, and replacing manual, device-by-device configuration with policy-driven routing intent managed through SmartConsole.
Partner POV | The Top Exposure Management Questions Security Leaders Ask: Insights from Customer Conversations
Explore how Check Point Exposure Management empowers security leaders to prioritize and remediate risks effectively. By correlating vulnerability data, validating exploitability, and reducing false positives, it streamlines response actions and enhances executive reporting. Discover strategies to manage cyber exposure and safeguard your organization with actionable insights and proven success rates.
What's New in R82.20, Part 1: AI Security Lands Directly on the Check Point Firewall
R82.20 revolutionizes Check Point Firewalls with AI Security, enhancing visibility and protection against AI threats. It introduces Workforce AI Security and AI Agent Security, offering granular control without new infrastructure. This release also supports SNORT 3.x, IPv6 DNS Trap, and provides a transparent TLS Inspection Block page, ensuring robust, native security.
Partner POV | Check Point Lays the Groundwork for the Future of AI Factory Security with NVIDIA
AI factories demand security that lives inside the infrastructure, not around it. NVIDIA's BlueField-4 and DOCA innovations, paired with Check Point's AI Factory Firewall, embed visibility and protection directly into the data path—securing autonomous, distributed AI operations at their source.