It was a private discussion with Fredrik Harrysson (Parity CTO) early July in Berlin. It sounds like plans have changed since then.
This is incorrect. Validators do not have to be VDF evaluators. Also, VDF evaluators do not need to be validators. That is, the two roles are completely decoupled.
VDF ASICs are radically different to mining ASICs in pretty much all respects. I hope education will allow the community to be well informed regardless of bad feelings about mining ASICs. A few points worth mentioning:
- Sequential vs parallel: VDFs are inherently sequential, i.e. massive parallelism does not help at all (the exact opposite to PoW mining). @djrtwo wrote a post about this. To give you an idea, the optimal parallel time circuit for modular multiplication (the basis for the RSA-based VDF we are considering) likely only uses a few mm^2 of die area (e.g. 3 mm^2).
- Power consumption: Current estimates suggest that a VDF evaluation would consume about 10 Watts. Assuming we have 10,000 VDF evaluations at any point in time (I am expecting only ~1,000, but let’s be conservative) that would amount to 0.1 MW. Compare this to today’s Ethereum mining which is about 2.3 GW (23,000 times more power intensive).
- Security margins: From the point of view of the VDF-based randomness beacon, the protocol can bake in a conservative security margin in terms of the speed advantage an attacker can have without getting any influence over the randomness. This is the A_{max} parameter defined here, and it will be carefully chosen.
- In-protocol rewards: For the VDF-based randomness beacon to function smoothly, I estimate in-protocol rewards to be ~$5K per day. Over a decade that corresponds to $18.5m. This is too low for a rational actor to build a somewhat faster proprietary ASIC to grab the in-protocol rewards. This is especially true if a ~$20m state-of-the-art commodity VDF ASIC is built in the first place. (In terms of the upfront R&D costs, the Ethereum Foundation is looking to pool funds with Filecoin and others.)
There is no rush to deliver the hardware. The Ethereum 2.0 roadmap (notably the first few phases) is unchanged with regards to the readiness of VDF hardware. Phase 0 (the beacon chain) is the short-term priority, and that will go ahead without any VDF. The VDF upgrade can come in phase 1 (sharding data layer), phase 2 (sharding state layer), or even later phases.
In regards to hardware manufacturing, this will be outsourced. We are in discussions for example with Obelisk Launchpad that can deliver a turn-key solution.