21 is widely expected for 2026. Harvard has already demonstrated 96 logical qubits. We only need about 15-20 to factor 21. So the hardware is big enough, error correction proven to scale, waiting for the accuracy of the gates to reach the point where it can handle a circuit of around 2,400 gates deep. COde word for that number of gates is deep logical circuits. I'd honesty not be surprised if a lab in china has already done it, they have orders of magnitude more funding than any lab in the US and typically do not announce results. But it's almost certain to happen in the US in 2026, be it harvard, quantiniuum, atom, or whomever else.