Quantum computing in 2024: Hype vs. Reality – still waiting for the 'quantum advantage'?

09 Tem 16:58

I still remember when IBM first rolled out the 1121-qubit Condor in 2023—everyone was hyping “quantum advantage” like it was tomorrow. Two years later, I’m running the same Grover demo on their updated 133-qubit Heron and spending half my day hunting for compiler knobs that actually silence the crosstalk. Bottom line: NISQ devices aren’t ready for prime time; they’re strictly a “trial-and-error” playground at this stage. So where can we actually chip away at real-world problems right now? My team switched gears last year and started wrapping quantum kernels inside classical HPC pipelines for chemistry simulations. We use a lightweight variational eigensolver on a 72-qubit Braket circuit at runtime, but offload the heavy lifting to a GPU cluster that handles the error mitigation and post-processing. This hybrid model cuts our compute time by ~40 % compared to classical DFT for certain small-molecule benchmarks—it’s not a “quantum advantage” in the strict sense, but it already saves us weeks per design cycle. If you’re itching to experiment without waiting for fault tolerance, pick a problem that (a) fits ≤127 qubits after transpilation and (b) tolerates ~1 % error rates. Quantum annealing on D-Wave for logistics optimisation fits that sweet spot; we’ve cut last-mile delivery costs by 8 % in a pilot route without touching the noise floor. The rule of thumb: if the classical competitor runs overnight, you’re still playing. If it finishes in seconds—and your quantum job finishes in milliseconds—then you’re seeing incremental leverage, not supremacy.