Cloud technology stacks and cost structures have gotten too complex.
Some workloads need to re-patriate to on-premise cloud architectures.
Andreeson-Horowitz asserts that Enterprise values would be collectively $1Trillion dollars higher if they repatriated their workloads from cloud providers. The claim is that repatriation is worth the work involved, and "Repatriation results in one-third to one-half the cost of running equivalent workloads in the cloud"
Venturebeat published an opinion piece disputing this notion. That cloud spend has an ROI to it - it enables the business to be more agile and take advantage of revenue blooms.
Workloads that were "shifted and lifted" to the public cloud don't leverage what the opportunity of the "cloud" provides. Not all workloads can or should be re-written to be a service model. Persistent traditional workloads, of which there are a ton, are good candidates to repatriate to a simpler, more cost-effective solution.
Regardless of where the workloads run the whole "cloud" thing has gotten too complex. Today's "cloud" requires many SKUs from multiple vendors, a variety of scarce human resources and complex billing. Traditional HCI and cloud stacks are not much better - and not very scalable.