The qubit drop that mostly isn't

The number falling is real. In 2019 the going estimate for breaking RSA-2048, the math behind most internet encryption, was around 20 million physical qubits. Google's team cut it under a million last year. Now Iceberg Quantum, a Sydney startup, puts it under 100,000 — roughly 98,000 for a month-long run, in the same range as machines a few labs already operate.

Inverting the matrix of a graph cannot realistically be achieved within the decoherence time. — a quantum decoder paper, on the missing piece

Then you check what the savings cost. Iceberg swapped the usual error-correcting scheme for a denser one that packs far more logical work into each physical qubit. But that density buys the lower qubit count by stretching the job out: measure the two designs in the unit that actually bills a quantum computer — qubits multiplied by days of runtime — and Iceberg's best case beats the prior surface-code estimate by about 1.5 times, not the ten-fold the headline implies. The work didn't get easier. It got rearranged into a longer, harder-to-sustain run.

And the run assumes a part nobody has built. The denser code needs a decoder that corrects errors in real time, every ten millionths of a second, across qubits wired to distant neighbours — and the paper says flatly that building it is out of scope. The largest code of this family ever run on real hardware is about seventeen times smaller. So the threshold that security teams race to outrun did drop on paper, while the thing standing between us and a codebreaker — the hardware, the decoder, a month of unbroken operation — moved not at all.