Gimme da (expensive) power

Although Fully Homomorphic Encryption makes things possible which otherwise would not be, it comes at a steep cost. Above, we can see charts indicating the additional compute power and memory resources required to operate on FHE-encrypted machine-learning models—roughly 40-50 times the compute and 10-20 times the RAM that would be required to do the same work on unencrypted models.

Source: IBM completes successful field trials on Fully Homomorphic Encryption | Ars Technica

Acquiring and maintaining this much computing power for FHE workloads is fucken expensive and that’s before you even start thinking about energy requirements for running this hardware and then cooling it.

Intel and AMD will be chomping at the bit to make us all buy new hardware though.