AI Development Navigates The Latency Sensitivity Spectrum: Training Allows For Slow Processing, But Real-Time Tasks Require Lightning-Fast Inference

 

Latency sensitivity in AI processes varies significantly between training and inference. Training operations, which involve processing large datasets over extended periods, are generally very tolerant of high latency. This tolerance allows training tasks to be performed with minimal concern for immediate responsiveness.

Wes Cummins, the CEO of Applied Digital joins David Liggitt, the Founder and CEO of datacenterHawk to talk about the spectrum of latency sensitivity within AI inference tasks. Mission-critical inference applications require ultra-low latency and high reliability, often needing to operate in cloud regions with five-nines reliability. Conversely, batch inference tasks, such as those involving generative AI for text-to-image or text-to-video conversions, can afford much higher latency. Chatbots and similar applications fall somewhere in between, with reasonable tolerance for latency variations.

Recent Episodes

AI infrastructure is evolving at breakneck speed, and the real challenge is no longer just designing next-generation data centers—it’s executing them at scale. As demand for AI-ready facilities grows, operators must adapt to immense increases in power density, new cooling technologies, and unconventional deployment locations. Power density requirements for AI workloads are pushing the…

As AI infrastructure spreads beyond tech hubs and into America’s heartland, companies face a new imperative: not just to build facilities—but to build trust, local partnerships, and long-term value for the communities that host them. In Ellendale, North Dakota, Applied Digital’s Polaris Forge 1 campus has become a case study in what rural revitalization…

As demand for artificial intelligence continues to soar, the AI infrastructure needed to power it is scaling just as rapidly. A 2024 report from the International Data Corporation (IDC) forecasts that global spending on AI infrastructure will exceed $200 billion by 2028, driven by an explosion in compute-heavy applications like large language models and…