The Quantum Computer Dream is Falling Apart

Quantum computing’s promise is running into a double bind: researchers are making steady progress on the hardware and error correction, yet practical “quantum advantage” for real problems keeps slipping away. Recent results show that tasks once used to justify quantum machines can be matched—or even outperformed—by conventional computing, while broader analyses argue that the most celebrated quantum use cases have not delivered evidence of a purely quantum speedup, even for small problem sizes.

A concrete example involves simulating a biologically important molecule known as FIMO co-actor, which plays a role in bacterial nitrogen enrichment in soil. The standard pitch is that quantum computers could compute the molecule’s properties—specifically the ground state energy—because conventional computers struggle with the underlying quantum equations. But a Caltech team reported in a new preprint that a conventional computer cluster can calculate the ground state energy with “stunning precision.” The significance isn’t that quantum computers are useless; it’s that the supposed path to a near-term, defensible advantage is harder than expected.

The skepticism extends beyond chemistry. The traveling salesman problem—often treated as a flagship benchmark for quantum approaches—has faced repeated attempts to recast it into a form where quantum hardware would outperform classical methods. A new paper reviews two decades of such efforts and concludes there’s little reason for optimism about purely quantum approaches delivering advantages. The authors argue for sticking with classical or hybrid classical-quantum systems instead, framing the key question as whether there’s evidence that fully quantum methods can solve even small traveling salesman instances better than hybrid strategies.

At the same time, the field’s technical trajectory is not collapsing. Error correction has been demonstrated on small circuits, qubit quality and precision have improved, and headlines have begun to use “transistor moment” language—an analogy to how microchips took off once transistors became cheaper and smaller. But the cost structure for quantum machines is different. Quantum systems rely on demanding cryogenic cooling and noise buffering, and those requirements don’t automatically get cheaper as more qubits are added.

There’s also an energy problem that complicates the “faster” narrative. A recent estimate by Olivier Rati suggests that once error-corrected quantum computers are large enough to do genuinely useful calculations, they could consume power on the scale of entire supercomputing clusters—potentially more depending on design. Even if quantum algorithms offer speedups for specific subroutines, achieving accurate results typically requires repeating computations many times, stretching runtimes and increasing the total energy and cost.

In short, quantum computing’s engineering progress is real, but the route to economical, demonstrable advantage is narrowing—especially when conventional clusters can already deliver high-precision results on some of the most persuasive targets.