Applied Digital is Scaling Up Infrastructure to Handle Growing GPU Deployment Needs

 

During a transformative period, Applied Digital experienced a significant increase in demand for large-scale GPU deployment. This demand required a focus on power density, as the networking within data centers, particularly for InfiniBand, necessitated that servers be located close together. Each server consumed 10.2 kilowatts of power, and the optimal performance was achieved when multiple servers were placed within 30 meters of the network core.

Wes Cummins, the CEO at Applied Digital in a conversation with David Liggitt , the Founder and CEO at datacenterHawk highlighted the challenge of finding large amounts of colocation (colo) space that could support interconnected GPUs. Despite securing some colo space from third parties to meet cloud demand for GPU deployment, it became clear that more extensive infrastructure was needed. Consequently, Wes instructed the real estate team to shift from planning 5 to 10-megawatt sites to constructing facilities with hundreds of megawatts, marking a significant strategic shift for the company.

Recent Episodes

Enterprise AI is advancing faster than most companies can govern it. Behind the scenes, AI systems are already influencing decisions tied to revenue, operations, compliance, customer outcomes, and risk — yet many organizations still lack a clear way to measure, explain, or oversee what those systems are doing. That is the gap TheAIAudit was…

Healthcare is being pushed to modernize faster than ever, as AI tools, virtual care, and digital patient experiences shift from innovation to expectation. Recent survey data from McKinsey & Company indicates that about half of U.S. healthcare leaders say their organizations have already put generative AI into practice, underscoring how quickly the technology is…

Artificial intelligence has already moved beyond the hype cycle and into the day-to-day reality of business operations. Companies across industries are rushing to integrate AI into their workflows, but many are running into the same challenge: it’s relatively easy to build something that works in a demo, and much harder to make it reliable…