Why does my randomness beacon assume no one has built a faster machine?

Randomness beacons used for validator selection, NFT minting, and DeFi parameter sampling rely on verifiable delay functions whose security guarantee rests on an assumption about how long sequential computation takes. If a well-funded adversary builds custom ASIC hardware that evaluates the VDF faster than the honest deadline, they can compute the beacon output in advance and bias any protocol that consumes it. A paper accepted to ACM CCS 2026 formalised this as an economic attack with a rational optimal-stopping structure and showed that several deployed parameterisations fall below the threshold needed to deter a hardware-equipped attacker during high-reward events like MEV spikes. No deployed beacon today publishes a hardware threat model or adjusts its delay parameter against observed ASIC progress.