Nvidia's Partner Highlights Banks' Demand for On-Premises AI Solutions
Perplexity CEO Aravind Srinivas highlighted a growing trend among major banks like Morgan Stanley and JPMorgan to adopt on-premises, air-gapped AI systems. These institutions prefer local hardware such as NVIDIA’s DGX Spark to maintain control over their data and prevent intellectual property from leaking into external labs. This shift reflects concerns around privacy and security in the financial sector. NVIDIA's CFO Colette Kress noted that trailing 12-month on-prem revenue reached $8 billion in automotive alone, with another $7 billion coming from financial services, manufacturing, and healthcare combined. The company anticipates non-hyperscaler categories like sovereign AI and air-gapped data centers will account for about half of its data center business. Srinivas also emphasized the limitations of centralized cloud computing for widespread AI adoption due to power and memory constraints. He advocated a hybrid model that leverages local hardware alongside cloud resources, tapping into untapped computational potential in offices and homes.
Topics: AI data centers hybrid AI infrastructure NVIDIA on-premises computing