Is Fully Homomorphic Encryption - cryptography's "holy grail" - inching closer to mainstream use?

The idea that we could run computational operations on encrypted data without ever needing to decrypt was first proposed in the 1970s and named “fully homomorphic encryption” (FHE). It’s the “holy grail” of cryptography but it took thirty years before a researcher named Craig Gentry was able to demonstrate that it could actually be done.

The math behind FHE is hard. Getting the hardware for it right has been too. FHE cipher text is big: running very large numbers that must be computed with precision is resource-intensive, and prospective enterprise buyers can’t live with tremendous latency, stymieing adoption.

In March 2026, FHE got another step closer to the possibility of mainstream adoption, after Intel showed the fruits of its five year programme to build dedicated FHE hardware, demonstrating its Heracles chip, able to compute with FHE around 5,000 times faster than Intel’s best CPUs, at a conference. 

Flavio Bergamaschi, a committee member of the Homomorphic Encryption (HE) Standardisation Consortium (HESC), told The Stack that Heracles addresses the longstanding bottleneck for the encryption “holy grail”, namely, the amount of data, computation and movement that it involves.