Quantum computing, without the hype: what is actually real
Google's Willow headlines, why qubits are fragile, harvest-now-decrypt-later, and the post-quantum encryption already rolling into your apps.
In December 2024, Google announced a quantum chip called Willow and claimed it solved a benchmark problem in five minutes that would take the fastest supercomputer ten septillion years — longer than the universe has existed, by a lot. The headlines wrote themselves. So did the confusion, because the benchmark was a task designed for quantum computers to be good at and useless for anything else, and the same announcement's real news was something far less quotable: error rates that finally fell as the chip got bigger, instead of rising.
Quantum computing lives permanently in that gap — between headlines that sound like magic and progress that is real but technical. Here is the honest version, including the part that already affects your passwords.
What is a quantum computer, without the mysticism?
A regular computer works with bits, each definitely 0 or definitely 1. A quantum computer uses qubits, which exploit two genuine features of quantum physics: a qubit can exist in a blend of 0 and 1 until measured, and qubits can be entangled so their fates link together. The result is a machine that can, for certain narrow problems, explore an enormous space of possibilities in a way no normal computer can match.
The catch is the word narrow. A quantum computer is not a faster laptop; on email, games, and spreadsheets it would be comically worse. It is a specialized instrument for problems with the right mathematical shape — simulating molecules for drugs and batteries, certain optimization and materials problems, and, infamously, breaking the encryption the internet runs on.
Why is it taking so long?
Because qubits are the divas of physics. They hold their delicate quantum state only while perfectly isolated — many designs run within a fraction of a degree of absolute zero, colder than deep space — and the slightest vibration or stray field corrupts them. Errors pile up so fast that useful computation drowns in noise.
The field's whole strategy is error correction: weave many fragile physical qubits into one reliable logical qubit. For decades that was theory, because adding qubits added more noise than it removed. The reason Willow mattered — and why the demonstrations that followed from IBM, Quantinuum, and others mattered — is that the trade finally started paying: bigger groups of physical qubits produced better logical qubits, not worse. That is the "it actually scales" moment the field had waited thirty years for. What remains is brutal engineering: useful machines likely need thousands of logical qubits, meaning millions of physical ones. Today's best machines have a handful of logical qubits and physical counts in the hundreds or low thousands. The gap is measured in years, plural, and nobody serious will say exactly how many.
Then why is your bank already reacting?
Because of a threat with a memorable name: harvest now, decrypt later. The encryption protecting most internet traffic relies on math problems ordinary computers cannot crack — but a large quantum computer running Shor's algorithm eventually could. So a patient adversary can record encrypted traffic today and simply wait to unlock it in the future. Anything that must stay secret for decades — state secrets, health records, financial archives — is effectively already exposed to a machine that does not exist yet.
That is why the response started before the threat: in August 2024, the US standards body NIST finalized post-quantum cryptography — new encryption built on math believed hard even for quantum machines — and the migration is quietly underway in your browser, your messaging apps, and government systems. It is one of the rare cases of civilization patching a roof decades before the storm, and it is the only part of quantum computing that touches your life today.
What should you actually believe?
A rough filter: distrust any claim that quantum computers will soon speed up everything, cure all diseases, or that they broke encryption last Tuesday — periodic papers claiming imminent code-breaking have so far collapsed under scrutiny. Trust the boring milestones: error rates falling, logical qubits multiplying, chemistry simulations inching past what classical computers can check.
The truthful summary is that quantum computing in the mid-2020s looks like classical computing in the 1950s: the physics works, the machines are room-sized and temperamental, the timeline is uncertain, and the people building them are simultaneously overpromising the decade and underestimating the century. The internet's encryption is already being replaced because of it. Everything else is a work in progress — real, slow, and nothing like the headlines.