In October 2019, Google published a paper in Nature announcing that its 53-qubit Sycamore processor had completed a computation in 200 seconds that would take the world’s fastest classical supercomputer roughly 10,000 years. The phrase “quantum supremacy” entered mainstream coverage overnight. The reality behind the claims is more nuanced — and more interesting.
## Google Sycamore (2019)
Sycamore’s task was random circuit sampling: draw a sample from the output distribution of a random quantum circuit. The task has no direct practical use, but it was engineered to be maximally hard for classical simulation.
IBM immediately disputed the 10,000-year estimate, arguing that classical algorithms using disk storage could complete the task in 2.5 days. Subsequent algorithmic improvements reduced the classical estimate further. Google’s result proved that the hardware worked correctly and that quantum systems could outperform classical ones on at least this specific task — an important engineering milestone, even if the supremacy gap has since narrowed. See [Google’s original paper](https://www.nature.com/articles/s41586-019-1666-5).
## China’s Jiuzhang (2020, 2021)
The team led by Jian-Wei Pan at USTC built a photonic quantum computer called Jiuzhang that performed Gaussian boson sampling. The 2020 version (50 photons, 76 detectors) was reported to be a quadrillion times faster than classical computers on this task. The 2021 upgrade, Jiuzhang 2.0 (113 photons), extended the claimed advantage further.
Gaussian boson sampling has potential connections to quantum chemistry and graph optimization, making it slightly more relevant to applications than random circuit sampling — though a practical quantum speedup for real problems remains distant.
## Xanadu Borealis (2022)
Xanadu’s Borealis processor used 216 time-multiplexed optical modes to demonstrate quantum advantage on a squeezed-light boson sampling task, completing the computation in 36 microseconds versus an estimated 9,000 years for classical simulation. A key feature: Borealis is programmable, not hard-coded to a single circuit. See the [Nature paper](https://www.nature.com/articles/s41586-022-04725-x).
## What These Milestones Actually Mean
All three demonstrations proved that quantum hardware can reach a regime where classical simulation becomes expensive for specific synthetic benchmarks. None proved a quantum speedup on a problem with direct economic or scientific value.
The current era of quantum computing — often called the NISQ (Noisy Intermediate-Scale Quantum) period — is defined by processors with 50–1,000 qubits, error rates too high for full error correction, and circuit depths limited by decoherence. Demonstrating practical quantum advantage, meaning a real speedup on a real problem, remains an open challenge.
The most promising near-term applications are quantum chemistry simulation (for drug and materials discovery), certain optimization problems, and quantum machine learning on specific model architectures. For a deeper look, see [Quantum Algorithms](https://sunqi.org/quantum-algorithms-en/) and [arxiv:quant-ph](https://arxiv.org/list/quant-ph/recent).
—
