Nvidia's Vera Rubin NVL72, unveiled at CES 2026, introduced rack-scale encryption, marking a significant advancement in enterprise AI security. The platform encrypts every bus across 72 GPUs, 36 CPUs, and the entire NVLink fabric, representing the first rack-scale platform to deliver confidential computing across CPU, GPU, and NVLink domains.

According to Louis Columbus, writing on January 12, 2026, this development fundamentally shifts the security paradigm for enterprise AI. Instead of relying on contractual trust with cloud providers to secure complex hybrid cloud configurations, organizations can now verify security cryptographically. This distinction is crucial, especially considering the increasing sophistication and speed of nation-state cyberattacks.

The need for enhanced security stems from the escalating costs associated with AI model training. Research from Epoch AI indicates that frontier training costs have grown at an annual rate of 2.4x since 2016. This exponential growth suggests that billion-dollar training runs could become a reality in the near future. However, the infrastructure protecting these substantial investments remains vulnerable in many deployments. Security budgets allocated to protect frontier training models are struggling to keep pace with the rapid advancements in model training, leaving more models exposed to threats as existing security approaches prove inadequate.

The Vera Rubin NVL72 aims to address this growing security gap by providing a comprehensive encryption solution at the rack scale. This approach ensures that sensitive data remains protected throughout the entire computing process, from data input to model output. The implications of this technology extend beyond mere data protection; it enables organizations to conduct AI research and development with greater confidence, knowing that their intellectual property is secure from unauthorized access.

The introduction of rack-scale encryption represents a critical step forward in securing the future of AI. As AI models become increasingly complex and valuable, the need for robust security measures will only intensify. Nvidia's Vera Rubin NVL72 signals a turning point in enterprise AI security, offering a new level of protection against evolving cyber threats. The industry will be watching closely to see how this technology is adopted and how it shapes the future of AI security.