> What must be understood is that in all these constructions, each circuit gate must map to an instance of Gentry's fully homomorphic encryption scheme, and at every clock cycle for the obfuscated circuit, all gates must be processed, regardless of whether they are "active" or not in the circuit (this is a big part of why the obfuscation theoretically works: it does not reveal active gates, by always making them all active). This[2] article gives performance result: on a rather big PC, we are up for minutes of computation. That's for each gate in the obfuscated circuit, and for each clock cycle.
> There are millions or even probably billions of gates in the envisioned circuit, given the setup of functional encryption: the "obfuscated circuit" must include asymmetric decryption and validation of zero-knowledge proofs. So now we are talking about using all the computers currently running on Earth, all collaborating on the computation, and they might make non-negligible progress towards running one instance of functional encryption within a few centuries.