{"id":72849,"date":"2025-09-05T08:00:00","date_gmt":"2025-09-05T15:00:00","guid":{"rendered":"https:\/\/www.salesforce.com\/?p=72849"},"modified":"2025-09-08T16:42:08","modified_gmt":"2025-09-08T15:42:08","slug":"scaling-cloud-infrastructure-for-agentic-enterprise","status":"publish","type":"post","link":"https:\/\/www.salesforce.com\/uk\/news\/stories\/scaling-cloud-infrastructure-for-agentic-enterprise\/","title":{"rendered":"Why Agentic AI Requires a Smarter Cloud Infrastructure"},"content":{"rendered":"\n

In today’s fast-moving business landscape, our customers are fundamentally reimagining how they engage with their own customers, leveraging agentic AI<\/a> to create dynamic, responsive experiences.<\/p>\n\n\n\n

This transformation, however, isn’t without its challenges. These AI systems must deliver on availability, performance, and scale, especially during peak and critical events. For enterprises that provide essential AI-powered services, the reliability of their systems is non-negotiable. As they embark on their journey to become an agentic enterprise, they require a trusted, robust infrastructure that can handle the unpredictable demands of AI, ensuring their business operates without a hitch. This is why our focus on engineering for AI scale is more critical than ever.<\/p>\n\n\n\n

\n

It\u2019s not simply about throwing more compute at the problem; it\u2019s about building a smarter, more dynamic, and inherently resilient foundation.<\/p>\n<\/blockquote>\n\n\n\n

Beneath the surface of agentic AI systems lies a critical, often underestimated, challenge: the underlying cloud infrastructure. It\u2019s not simply about throwing more compute at the problem; it\u2019s about building a smarter, more dynamic, and inherently resilient foundation.<\/p>\n\n\n\n

The Unique Workload Demands of Agentic AI<\/strong><\/h2>\n\n\n\n

Unlike traditional enterprise applications with predictable usage patterns, agentic AI workloads are inherently unpredictable. Along with increased reasoning capabilities<\/a> for more precise outcomes, they demand bursts of intense compute power, often requiring specialized hardware like GPUs, alongside rapid data ingress and egress. This burstiness, coupled with the need for ultra-low latency and high throughput, pushes conventional cloud infrastructure to its limits. Reactive scaling mechanisms, designed for more stable demand, fall short when faced with the volatile, diverse resource needs of intelligent agents.<\/p>\n\n\n\n

\n

To truly unlock the potential of agentic AI, we must move beyond mere elasticity and embrace an infrastructure that is intelligent, predictive, and highly efficient.<\/p>\n<\/blockquote>\n\n\n\n

To truly unlock the potential of agentic AI, we must move beyond mere elasticity and embrace an infrastructure that is intelligent, predictive, and highly efficient. We need engineering systems that can anticipate demand, optimize resource allocation across a heterogeneous fleet, and enable rapid iteration for continuous improvement. This requires a fundamental reimagining of how we build and manage our cloud environments.<\/p>\n\n\n\n

Engineering for AI Scale: Key Innovations<\/strong><\/h2>\n\n\n\n

At Salesforce, we’re deeply invested in evolving our cloud infrastructure to meet these exacting demands. This includes developing and deploying innovations that directly address the unique characteristics of AI workloads:<\/p>\n\n\n\n