ReservarTienda

Noticias Cuánticas: lunes, 29 de junio de 2026

Three stories this week, and they rhyme: the field is being held to harder numbers, by harder critics. Start with the deadline. The U.S. Department of Energy announced 'Quantum Genesis,' an initiative to stand up a fault-tolerant, scientifically useful quantum computer by 2028 — and unlike most roadmaps, it comes with a brutal spec. The DOE 'Q Competition' wants systems with 150 to 250 logical qubits (the error-corrected kind, each one stitched together from many noisy physical qubits) running circuits at a logical error rate of 10⁻⁸ — one slip in a hundred million operations. For comparison, the best logical demonstrations earlier this year were hovering around breakeven, the line where a logical qubit barely outlives the physical ones underneath it. So 10⁻⁸ by 2028 is either a forcing function or a setup for a missed target; either way the spec deadline is this September, which is when we'll find out whether anyone actually signs up. Second, an architecture result that makes that goal look less impossible. QuEra and Los Alamos published, in PRX Quantum, a neutral-atom design they call transversal STAR that cuts the physical-qubit cost of useful quantum simulation by orders of magnitude — roughly 1,500 to 3,000 physical qubits for problems that conventional fault-tolerant schemes would need hundreds of thousands for, at about a 250x speedup. The trick is co-design: instead of compiling a generic circuit, they line up the symmetries of the physics problem with the structure of the error-correcting code and skip magic-state distillation (the expensive ritual that normally dominates the qubit budget) altogether. It only works for structured simulation problems, not arbitrary computation — but structured simulation is exactly what the DOE wants the machine for, so the two stories slot together. Third, the skeptic. Henry Legg, a physicist at St Andrews, published a formal challenge in Nature's 'Matters Arising' section arguing that Microsoft still hasn't shown a topological qubit on its Majorana 2 chip — that the signatures it points to could be measurement noise or ordinary quantum-dot effects, and that some of the tune-up routines and analysis code don't hold up. Microsoft's Chetan Nayak replied that the company 'stands by our results and our roadmap.' We don't know who's right, and that's the point: this is how the field is supposed to work — claims published, claims challenged, in the same journal, in the open. The through-line: a national lab setting a hard error-rate target, an architecture that makes the target cheaper to reach, and a respected critic forcing a giant to defend its data. None of it is a new qubit-count record. All of it is the quiet, unglamorous machinery of a field growing up.