tumbleweed it is, but the direction might be wrong
What we see here is indeed a move away from vSphere, but into different directions.
As a benefit of moving some workloads into the cloud, the size and managability needs for the on-premises part of the mix tend to be reduced, often allowing a move away from the full vSphere product with huge license savings.
With many users we know of, this is picked up by either ESXi (if the remaining workloads are small enough and no vMotion is needed) to leverage existing skills and kit, or more often by KVM introduced by a next generation of admins, who are natural with command line Linux administration and run virt-manager from their Ubuntu workstations, Macs via XQuarz or even Windows via Xming. These bring bcache, DRBD and friends to the table, providing quite a bang for the buck for the investment of a set of skills, that are much more easily available today than they were a decade ago.
Hyper-V (in all its flavours) doesn't see many vSphere converts here: The licensing model is not so different from vSphere as to be convincing, and asking an Admin to run Windows 10 as a hard dependency to administer his VMs is a very hard sell.