John Preskill, one of the leading voices in quantum computing, recently wrote about something that might sound like bad news for the field. A classical algorithm managed to solve a chemistry problem - the FeMo-cofactor structure - that many had assumed would require quantum computers. But his takeaway isn't what you might expect.
In his reflection on the result, Preskill argues this isn't a setback for quantum computing. It's a clarification about where quantum advantage actually lies.
The FeMo-Cofactor Problem
The FeMo-cofactor is a complex molecular structure at the heart of nitrogenase, the enzyme that converts atmospheric nitrogen into ammonia in bacteria. Understanding its electronic structure has been a longstanding challenge in computational chemistry.
For years, this problem was held up as a prime candidate for quantum computers. The argument went that simulating quantum systems - like the electrons in this cofactor - should naturally suit quantum hardware. Classical computers would struggle with the exponential complexity.
Then researchers developed a classical algorithm that solved it to chemical accuracy. The quantum advantage everyone assumed was there... wasn't.
What Actually Happened
Preskill's insight is that the breakthrough wasn't about raw computational power. It was about clever state preparation. The researchers found a way to construct a good initial guess for the quantum state, which allowed classical algorithms to converge efficiently to the answer.
In simpler terms - they figured out where to start looking, which made finding the answer much easier. That's different from the fundamental question of whether quantum computers can solve problems classical computers fundamentally cannot.
This distinction matters. The FeMo-cofactor result shows that some problems we thought required quantum computers might yield to better classical algorithms and smarter problem setup. But it doesn't mean quantum computers have no advantage elsewhere.
Where Quantum Advantage Really Lies
Preskill points to problems where we can't cheat with clever state preparation - situations where you genuinely need to explore a vast quantum state space without shortcuts. These still exist, even if the FeMo-cofactor wasn't one of them.
The challenge for the quantum computing field is being honest about which problems fall into which category. Overpromising on applications that might have classical solutions undermines credibility. But abandoning the technology because one benchmark fell to classical methods would be equally misguided.
What this really reveals is how much we still don't know about the boundary between quantum and classical advantage. That boundary keeps shifting as researchers develop better algorithms on both sides.
The Scientific Process at Work
There's something refreshing about Preskill's response to this result. He doesn't downplay the classical achievement or insist quantum computers would still be better somehow. He acknowledges the reality and uses it to refine understanding of where quantum advantage actually applies.
This is how science should work. A hypothesis about where quantum computers would excel gets tested. Reality provides data. Understanding improves. The field moves forward with clearer boundaries and more honest expectations.
For anyone watching the quantum computing space, this is a useful reminder to be sceptical of broad claims about quantum advantage. The real question isn't "can quantum computers solve hard problems" but rather "which specific problems require quantum approaches, and which might yield to better classical methods."
What This Means for Development
The practical implication is that quantum computing research needs to be more precise about target applications. Not every chemistry problem requires quantum hardware. Not every optimisation challenge needs quantum annealing. Some problems will, but identifying them accurately matters more than claiming broad superiority.
For businesses considering quantum computing investments, this clarifies the timeline. If your specific use case might be solvable classically with better algorithms, waiting for quantum hardware might not be the answer. Conversely, if your problem genuinely requires quantum approaches, the investment case becomes clearer.
The FeMo-cofactor result doesn't diminish quantum computing's potential. It sharpens our understanding of where that potential actually lies. That's progress, even if it looks like a step backward at first glance.