Garnet Chan never bought the idea that certain chemical mysteries demanded quantum computers. For years the Caltech chemist pushed back against claims that systems like the iron-molybdenum cofactor at the heart of nitrogenase lay beyond classical reach. Last month his team delivered the proof.
They computed the ground-state energy of the FeMo-co cluster to chemical accuracy. All on ordinary machines. The Quanta Magazine report lays out how this decades-long effort upends assumptions that have driven millions in quantum hardware investment.
Nitrogenase turns atmospheric nitrogen into ammonia. Bacteria have done it for billions of years under ambient conditions. Industrial Haber-Bosch plants require high heat and pressure. Understanding the enzyme’s mechanism could reshape fertilizer production or catalyst design. Yet the electronic structure of its active site has resisted precise calculation.
The cofactor contains seven iron atoms with highly entangled electrons. Earlier estimates put the number of relevant electron configurations above 78,000. Many researchers figured only a quantum computer could handle the superposition and correlation. Microsoft researchers highlighted nitrogenase as a benchmark in 2011. A 2017 PNAS paper framed it as a test case for quantum advantage.
Chan disagreed. He suspected classical methods could suffice if applied cleverly. His group’s January 2026 preprint on arXiv demonstrated exactly that. They reached chemical accuracy for the 76-orbital, 152-qubit resting-state model without invoking quantum hardware.
How? Two complementary classical compression strategies. One starts from an educated guess at the dominant configurations and incrementally adjusts electrons. Larger changes contribute negligibly to the total energy, so those configurations can be safely ignored. The second decomposes the initial state into pieces that exchange limited information. Anything beyond a chosen threshold gets truncated.
Both approaches converged on the same ground-state energy. The numbers match experimental observations of the biological system. “I think it’s important to clarify that this is not an impossible task where you have to first build a quantum computer to say anything about the problem,” Chan told Quanta.
The result lands at a moment when classical simulations of quantum systems keep surprising the field. Only weeks earlier, researchers at the Flatiron Institute’s Center for Computational Quantum Physics overturned a 2025 claim of quantum supremacy in qubit dynamics. Joseph Tindall and Miles Stoudenmire adapted a belief-propagation algorithm from the 1980s to tensor networks. They ran initial calculations on a laptop using the ITensor library. Their Simons Foundation announcement notes the simulations matched the quantum results yet required no quantum device.
Tindall described tensor networks as “a zip file for the wave function.” The compression lets classical machines tackle entanglement that once looked prohibitive. Similar ideas underpin Chan’s success with FeMo-co. Decades of refinement in tensor methods, density matrix renormalization, and selective configuration interaction have expanded classical territory.
But. Not everyone sees the work as closing the book. Daniel Suess, commenting in Quanta, noted the calculation addresses only the resting state. “We’re not even close to achieving the holy grail of this. We’ve still just described the resting state. But the method is promising in that it suggests we can proceed with some confidence.”
James Whitfield raised a deeper point. Solving one specific molecular instance after 20 years of optimization does not guarantee transferable methods. “If we pick any optimization problem and you put 20 years into it, you can figure out that one system. But whether that solution is transferable? Questions like that won’t be answered by solving one instance of one molecular system.”
Chan acknowledges quantum computers will bring value. He would use one tomorrow for certain tasks. Dynamic simulations of reaction pathways, where time evolution matters, may still favor quantum approaches. Yet the new evidence suggests chemists need not wait for fault-tolerant machines to tackle hard ground-state problems.
His team’s paper, titled “Classical Solution of the FeMo-Cofactor Model to Chemical Accuracy and Its Implications,” appeared first on arXiv in January. Chan followed with a March post on the Quantum Frontiers blog explaining the misconceptions that had grown around the problem. The work builds on years of incremental advances in classical quantum chemistry algorithms.
Recent large-scale classical simulations reinforce the pattern. In March 2026 a Japanese team simulated quantum chemistry circuits on 1,024 GPUs, pushing past previous 40-qubit limits for benchmarking. Hybrid quantum-classical ideas continue to surface too. Researchers from IonQ and Microsoft proposed in IEEE Spectrum that quantum machines might generate training data for AI models that then run fast predictions on classical hardware.
Still the FeMo-co result stands out. It targets a system long sold as quantum territory. The classical victory does not kill interest in quantum chemistry simulation. It reframes the timeline and the expectations.
Scientists have chased accurate nitrogenase models since the enzyme’s structure came into focus. Early quantum resource estimates assumed exponential scaling would doom classical efforts. Improved heuristics changed the math. Chan’s group showed that with the right initial guess and systematic truncation, the exponential wall develops cracks.
The broader lesson stretches past one enzyme. Quantum computing advocates have pointed to chemistry as a near-term killer application. Fault-tolerant devices could, in theory, deliver exact energies and rates for catalyst screening or drug design. That promise helped justify billions in public and private funding.
Yet classical tools keep advancing in parallel. Tensor networks, machine learning potentials, and clever basis-set choices erode the cases once thought strictly quantum. The debate now centers on where the crossover happens and how soon.
Chan remains measured. “My main interest is in solving chemical problems. If classical computers are the right tool to do it, we should use them. I don’t see why we should wait for a fault-tolerant quantum computer to be built.”
His stance echoes a growing chorus. Recent workshops at Simons Berkeley and Aspen have explored quantum advantage for computational chemistry in an era when classical simulation keeps extending its reach. The FeMo-co calculation adds concrete data to those discussions.
Practical payoffs remain distant. Even perfect knowledge of the resting state does not automatically yield a room-temperature industrial catalyst. Biological systems evolved for their own purposes, not manufacturing efficiency. Still, the computational milestone removes one excuse for delay.
Researchers can now probe the full enzyme with greater confidence using classical resources. Extensions to reaction intermediates and dynamics lie ahead. Some of that work may eventually run on quantum hardware. Much of it probably will not.
The field has spent years preparing for quantum supremacy in chemistry. This result suggests parts of that supremacy may prove illusory. Classical machines, sharpened by mathematical insight, can deliver answers today.
And the nitrogenase story is far from over. But one central question has a clearer answer than before. For at least this benchmark, quantum computers are not required.
WebProNews is an iEntry Publication