Every field has a word it cannot shake off, and quantum computing’s was minted in 2012 by John Preskill, Caltech’s Richard P. Feynman Professor of Theoretical Physics. “Quantum supremacy”, as glossaries now define it, means the demonstration of a quantum computation that solves a problem which “cannot be solved classically in any realistic timeframe”, the technical bar being a “superpolynomial speedup over all possible classical approaches” to the same problem. It is a deliberately narrow claim, and the narrowness is the point: the term’s inventor is a theorist allergic to overclaiming, which the fine print of his own coinage makes clear.
The record of the man, on his own Caltech page, reads like a career spent moving toward the subject rather than starting in it. A theoretical physicist by training, his earlier work touched particle physics and cosmology, topological defects, the nonperturbative side of quantum field theory, and quantum questions around the early universe and black holes. From the mid-1990s the center of gravity moved to quantum information, computing and error correction, and the engineering worry that drives it shows in his stated interest in “protecting intricate quantum systems from decoherence and other sources” of error, keeping delicate quantum states alive against the noise that destroys them. Institutionally, he founded the Institute for Quantum Information in 2000 under a National Science Foundation initiative, and in 2011 that institute was folded into the broader Institute for Quantum Information and Matter, an NSF Physics Frontiers Center. He teaches quantum computation as Physics 219, blogs for the IQIM’s Quantum Frontiers, and has sent 85 former postdocs onward to faculty positions.
Under the hood, the supremacy term is stricter and humbler than the headlines suggest. It carries no qualitative requirement, so it can apply to demonstrations of problems with no real-world applicability, and to results achieved with neither quantum error correction nor error mitigation applied. It asks neither for commercial viability nor for fault-tolerant quantum computing, which is why such demonstrations count as academic endeavors first, the natural business of the Noisy Intermediate-Scale Quantum (NISQ) era. The argument for caring anyway is a confidence argument: solving classically intractable but useless problems builds trust that classically intractable and useful ones can be found.
The vocabulary has since grown appendages to cover what the industry actually ships. Supremacy, as the glossary draws the lines, means classical solutions are infeasible; “quantum advantage” means the quantum solution is faster or more accurate than classical rivals; “quantum utility” concedes there is no particular advantage, but it works. On qubit counts, a standard baseline for supremacy is “at least 50 qubits”, beyond what the most powerful simulators can reproduce, though harder problems need more. And the prize sketched for genuinely useful supremacy is a familiar wishlist: decryption of formerly secure data via Shor’s factoring algorithm, precise simulation of molecules that today can only be approximated, the analysis of absolutely massive datasets, and large combinatorial optimization.
The arc of Preskill’s own talk titles is the field’s diary. In 2015 he was asking “Is spacetime a quantum error-correcting code?”; by 2017, the Q2B keynote was “Quantum computing in the NISQ era and beyond”; in 2019 came “What’s next after quantum supremacy”; in 2020, a KITP lecture on “The ghost in the radiation: robust encodings of the black hole interior”; in 2023, “Crossing the quantum chasm: from NISQ to fault tolerance”; in 2024, “Beyond NISQ: the megaquop machine”; and in 2025, “Quantum computing in the second quantum century”. A Bell Prize session in 2024, a fortieth-anniversary Feynman lecture in 2018 and a 2021 oral history interview with David Zierler round out a paper trail that never sits still.
The technical reality is that the most famous experiment attached to the word, the Google-NASA-Oak Ridge National Laboratory demonstration, arrived surrounded by exactly the misunderstandings the coiner’s small print was designed to prevent, along with calls for patience before large-scale, error-corrected, genuinely useful machines. The detail that matters is that the definer has spent the decade since 2019 asking what comes after his own milestone, from the chasm to the megaquop to the second quantum century. As ever with a good theorist, the term was the beginning of the argument, not the end of it: a word built to be outrun, by an author who has obligingly kept running.
