Four Quantum Layers Moved In One Day: Hardware, Simulation, Funding, And Factory Software

Four parts of the quantum stack moved in the same 24 hours, and the parallel motion is the news.

Quantinuum's Helios trapped-ion machine reached 99.9995% logical fidelity — five errors per million operations — on a color-code error-correction layout, with an error rate 10x lower than the company's own September record, according to its release. Quantinuum frames that as roughly 100x lower error than the best published result on any other hardware modality, a Google superconducting number. The honest read: superconducting is the floor Quantinuum is beating, not a category it has eliminated.

The same day, NVIDIA and the Jülich Supercomputing Centre ran a 104-qubit fault-tolerant simulation on JUPITER, Europe's first exascale supercomputer, using 8,192 NVIDIA GH200 GPUs, announced just ahead of GTC Berlin in Berlin. The run modeled magic-state cultivation — preparing the "magic" states an error-corrected computer needs for non-Clifford operations — under realistic noise; the approach NVIDIA validated, the company says, needs 90% fewer physical qubits than previous methods. It's a simulation, not hardware, but the sandbox where the next error-correction tricks get tested.

Xanadu received provisional approval for up to $10M more under DARPA's Quantum Benchmarking Initiative — a one-year contract extension building on its Stage B work — for its photonic computing program.

Eyelit put its MES/MOM platform, the manufacturing-execution and operations-management software used in semiconductor fabs, live at another quantum chip factory. The least photogenic of the four, and arguably the one that decides whether the hardware news actually ships at scale.