On October 23, 2019, Google announced that its Sycamore quantum processor had performed a computation that no classical computer could feasibly reproduce, a milestone it called quantum supremacy. The result appeared the same day in the journal Nature. Sycamore is a chip of 54 superconducting qubits arranged in a two-dimensional grid, each connected to four neighbors, operated at temperatures near absolute zero.
The task was deliberately narrow: sampling the output of a random quantum circuit of roughly 1,000 gates. The point was not usefulness but hardness - finding the most likely bit-strings such a circuit produces becomes exponentially harder for a classical machine as the qubit count and circuit depth grow. Google reported that Sycamore did the job in about 200 seconds, and estimated that the world’s fastest supercomputer would need around 10,000 years to match it. The company framed this as the first experimental challenge to the extended Church-Turing thesis, the long-held assumption that a classical computer can efficiently simulate any reasonable model of computation.
The claim drew immediate pushback. IBM argued that with better use of disk storage a classical supercomputer could complete the same task in about 2.5 days rather than millennia, which would shrink the gap dramatically without erasing it. The episode set the tone for the field: headline supremacy demonstrations on contrived benchmarks, followed by classical algorithms clawing back ground, with the practical payoff still distant.
This entry sits in the AI library as a compute milestone rather than an AI result. Sycamore solved nothing that machine learning needs. But it is the clearest public marker that a genuinely new computing substrate has crossed from theory into the lab, and it anchors the long-shot hope behind quantum machine learning.