Fraunhofer Papers Question the Road to Quantum Advantage

The story often told about quantum advantage goes like this: a machine surpasses a classical computer, beats it at a practical task, and the industry has its proof. Two new publications from the Fraunhofer Institute for Applied Solid State Physics IAF reveal a different narrative. The threshold hasn’t been crossed for any real-world application, and most of the claims made about quantum advantage rest on assumptions that don’t hold up when faced with real-world molecules or large problems.

The idealized molecule problem nobody wants to account for

Quantum chemistry has a habit: much of the literature treats molecules as isolated systems, with no interaction with their surroundings, governed solely by unitary dynamics, and calculated using the Born-Oppenheimer approximation. However, none of these conditions are present in nature. Molecules leak energy, relax, and settle into thermal states due to interactions with their environment.

The Fraunhofer review, co-authored with HQS Quantum Simulations and ETH Zurich, suggests that dissipation should be considered a resource for stabilizing useful quantum states rather than as noise to suppress. This is a fundamental shift in the method of building algorithms. It also implies that demonstrations within idealized models tell us very little about performance under real-world conditions.

Note the collaboration: HQS sells quantum simulation software, so this paper promoting open-system methods aligns with their business interests. Though this doesn’t make the physics wrong, it does create an incentive to present the argument in a certain way.

When, why, and under what conditions

Co-author Dr. Florentin Reiter, head of the Quantum Systems business unit at Fraunhofer IAF, puts it this way:

“The exciting question is not just whether quantum computers can outperform classical computers, but when, why, and under what conditions. For chemistry, this means we should not only consider idealized, closed systems but also the open dynamics that are ubiquitous in nature.”

This is an inward-aiming correction. The field spent years asking whether quantum computers can outperform classical computers; Fraunhofer is asking under what conditions, a much harder question. It’s not a straightforward answer that can be provided by a single experiment. Instead, it requires specifying the physical regime where the advantage holds, and this specification tends to narrow the claim.

Scaling and the QAOA story

The second paper, authored by Vanessa Dehn, examines the Quantum Approximate Optimization Algorithm (QAOA) for combinatorial problems in finance, logistics, network planning, materials design, and machine learning. The central question is how computational cost changes as the problem grows, rather than whether the algorithm works on small instances.

Here’s the part that press releases gloss over. The paper reports that favorable scaling over classical algorithms may be possible for portfolio optimization within the examined problem sizes. However, the phrase “may be possible” carries significant weight. The claim is based on simulation results, not a hardware demonstration or asymptotic advantage. QAOA’s scaling behavior at large depth and large problem size remains uncertain due to the history of small-instance trends not extrapolating.

Technical nuance skipped by summaries

One detail is often omitted: QAOA on classical hardware can only be simulated for a limited number of qubits before the state vector becomes intractable to store. Every claim about scaling derived from simulation therefore extrapolates from a small, inherently limited sample.

The procedure for transferring optimized parameters from small problems to larger ones is the interesting aspect, but also the load-bearing assumption. If the transfer degrades faster than the model predicts, the projected advantage disappears. Simulation can indicate a trend, but it cannot guarantee one at the sizes where quantum machines would actually prove their value.

Fraunhofer would argue that this highlights the need for better benchmarks. A field relying on unrealistic demonstrations needs someone to redefine the standard. Dehn’s contribution captures the spirit of discipline needed:

“Small-scale demonstrations alone are not enough. The crucial question is what happens as a problem grows larger. That’s where it becomes clear whether an approach can remain relevant in the long term.”