I’m continuing a series of broad review papers on quantum computing after Disentangling quantum emulation and quantum simulation in January 2023, Is there a Moore’s law for quantum computing? in March 2023 and Perspective on superconducting qubit quantum computing published in EPJA in April 2023.

This time, I evaluate the state of the art of noisy intermediate scale quantum computers (NISQ) and what can be done with it or not, and how it can improve in the near future. I lay out some inconvenient truths: NISQ is not at all ready for prime time quantum computing despite all the fuss about “quantum computing being business ready”. Not only have we not yet reached any quantum computing advantage, but in many cases, even if it worked, the most common NISQ algorithms using a variational approach, have prohibitively long execution times particularly in the promising chemical simulations domain. Most documented gate-based NISQ algorithms use cases run with fewer than 20 qubits that can be emulated faster on a regular laptop costing less than $2K and even provide better results since they emulate perfect non noisy qubits. Also, there are significant inconsistencies between the criteria to reach some quantum advantage (>50 qubits, some computing depth) and the fidelities of the required physical qubits.

There is still hope to extract some value from NISQ quantum computers, mostly with analog quantum computers and with various other techniques related to the improvement of gate-based NISQ quantum computers, but probably within a rather narrow window, corresponding to the 100×100 (# qubits x computing depth) challenge proposed by IBM for its future Heron 133 qubits QPU.

You can download this paper on arXiv.

If you are interested by the advent of FTQC (scalable fault-tolerant quantum computing), my recent arXiv on Moore’s law in quantum computing is a good starter.

Thanks a lot to Alain Chancé (Molket), Vincent Elfving (Pasqal), Marco Fellous-Asiani (Centre of New Technologies University of Warsaw), Loïc Henriet (Pasqal), Michel Kurek (Multiverse), Jean-Baptiste Latre (Qualitative Computing), Joseph Mikael (EDF), Stéphane Requena (HQI/GENCI), Jean Senellart (Quandela), Simone Severini (AWS), Robert Whitney (CNRS LPMMC, QEI), Xavier Waintal (CEA IRIG) and Raja Yehia (ICFO) for their feedback on the paper, which doesn’t mean an endorsement of all its content.

I also recognize useful discussions with and insights from Alexia Auffèves (CNRS MajuLab, CQT, QEI), Cyrille Allouche (Atos/Eviden), Thomas Ayral (Atos/Eviden), Jerry Chow (IBM), Jay Gambetta (IBM), John Preskill (Caltech), and Pierre Perrot (Eureptech).

I wrote this paper from February to May 2023 and I am ready to correct errors and complement it for a subsequent release!