Signal A 60% to 70% cloud-to-hardware cost ratio is named as the repatriation tipping point

Summary

Deloitte offers a concrete threshold for deciding where AI workloads should run. For consistent, high-volume workloads, on-premises deployment may become cheaper than cloud services once cloud costs start to exceed 60% to 70% of the total cost of acquiring equivalent on-premises systems. At that point capital investment becomes more attractive than ongoing operating expense for predictable AI work. AI thought leader David Linthicum, interviewed for the chapter, says that with all else equal he would choose public cloud for its ease, scalability and elasticity. Once costs cross that ratio, however, he advises evaluating alternatives such as colocation and managed service providers. He presents the metric as a practical, quantifiable guide that replaces defaulting to cloud-first regardless of economics.