DecaQ says 108-qubit FeMoco phase estimation finished in 2.045 seconds

DecaQ.ai says it ran a FeMoco-derived quantum phase estimation workload on its DecaQuasar BB3 architecture with a 2.045-second median full-pipeline time. The result highlights a 108-logical-qubit target-state workload, cross-node entanglement and verified output, but it stops short of a full FeMoco ground-state energy calculation.

Why it matters: - FeMoco is a benchmark problem in quantum computational chemistry because it matters for nitrogen fixation research, catalyst design and fertilizer-related science. - DecaQ’s result is a software-and-architecture claim, not a hardware-qubit milestone, and it is framed around deterministic execution and verification on general-purpose computing hardware. - The reported workload combines non-Clifford preparation, entanglement and non-commuting Hamiltonian terms, which are the kinds of features that make quantum chemistry workloads hard to scale.

What happened: - DecaQ.ai announced results on September 24, 2026 for a FeMoco CAS(54e,54o)-derived quantum phase estimation workload executed on its DecaQuasar BB3 architecture. - The company says the benchmark recorded a median full-pipeline time of 2.045 seconds across five runs. - The result covers workload generation, execution and verification. - The announced workload used 108 logical target qubits.

The details: - The QPE-4M workload includes non-Clifford state preparation, cross-node entanglement and non-commuting Hamiltonian execution. - The scientific report records 270 state-preparation gates, including 54 T gates and a 54-edge cyclic CZ interaction layer. - The prepared state has support over 2^54 computational-basis components, with 54 electrons represented in every component. - The report records eight entangled pair links across the target-node partition and a node-partition Schmidt rank of 256. - The Hamiltonian comprises 162 generator groups and expands to 270 Pauli components. - The Hamiltonian includes 216 anticommuting component pairs. - The execution spans two target nodes and one QPE control node. - Forty Hamiltonian components cross the target-node partition. - Across 80 QPE precision rounds, the execution record reports 12,960 generator-group evaluations, 21,600 expanded-component evaluations and 3,200 cross-node component evaluations. - The exact recorded eigenphase is 27/524288, returned as an 80-bit binary string. - The execution record reports no dense global statevector allocation and zero materialized global amplitudes. - The benchmark also reports a median process-level execution time of 0.709 seconds across nine runs and a median in-process engine execution time of 0.116 seconds across 15 runs. - The publication package includes the scientific report, complete OpenQASM target-state preparation source and a timing appendix. - Artifact identifiers connect the benchmark to the workload documented in the report. - The announcement says the result concerns phase evaluation for the specified FeMoco-derived workload and does not report a full FeMoco ground-state energy calculation. - The OpenQASM attachment specifies target-state preparation, while the report separately describes the QPE-4M Hamiltonian execution.

Between the lines: - DecaQ is positioning the result as evidence that a digital-quantum-oracle approach can handle a chemistry-style workload with entanglement and non-Clifford structure without materializing a full statevector. - The emphasis on full-pipeline timing suggests the company wants to frame performance around a complete workflow, not just engine runtime. - The distinction between phase estimation and ground-state energy calculation matters because the announced result is a narrower benchmark than a full chemistry solution.

What's next: - DecaQ is likely to use this benchmark package to support follow-on claims about larger or more complex quantum algorithm workloads. - Any broader validation will depend on whether the company can reproduce similar results across additional chemistry problems and scaling regimes. - The current announcement leaves open how the architecture performs on end-to-end scientific tasks beyond phase evaluation.