Swiss Team Claims Quantum Chip Setup Generates Truly Perfect Random Numbers
ETH Zurich researchers say two superconducting qubits and 30 meters of microwave pipe can certifiably eliminate bias from cryptographic key generation — forever.
Randomness is harder than it looks. Coins are biased. Dice are biased. Software pseudo-random number generators have shipped with enough predictability to compromise IoT devices at scale and, at one point, WhatsApp's key generation. The failure mode here is always the same: entropy sources that look random but aren't, and the downstream cryptographic systems that trust them anyway.
Researchers at ETH Zurich say they've closed that gap — at least in principle.
Their setup: two superconducting chips, each representing a single qubit, cooled to near absolute zero and connected by a 30-meter microwave guide held at the same temperature. Microwave photons traveling between the chips produce quantum entanglement. The output of that process runs through an algorithm designed to amplify randomness and strip out any residual bias. The team published the results in Nature under the title "Experimental randomness amplification."
Team leader Renato Renner put it flatly: the resulting bit sequence "is now really perfectly random, and we can even certify that." Co-leader Andreas Wallraff noted that even photon beam-splitter-based quantum RNGs — currently considered state of the art — aren't fully immune to systematic bias. This approach, they argue, is.
In practice, the immediate applications the team points to are cryptographic key generation and public randomness services for lotteries or blockchain consensus mechanisms. Both of those are genuinely high-value targets. Blockchain VRF implementations have been manipulated before. Lottery RNG audits are their own cottage industry of forensic distrust.
The caveat worth flagging: the researchers are not the first to claim perfect or near-perfect quantum randomness. The field has a history of results that hold up in lab conditions and degrade under real deployment constraints — temperature variation, timing side-channels, software integration bugs. One thing the post-mortem will say, when someone eventually finds a crack, is that the hardware was fine and the API wrapper wasn't.
Still, certifiable randomness with a formal proof attached is a different category of claim than "we think this is random enough." If the math holds at scale, this matters for anything building on cryptographic primitives.
Get the entropy source right before you worry about the key management policy.
