Quantum computing: What can it do and what cannot it do?

Starting from the most basic principles, gradually pushing forward to the current situation

You are likely to have seen headlines like 'Google quantum computer completes 10000 year supercomputing task in 3 minutes' or' China's Zuchong III is launched, 15 orders of magnitude faster than supercomputing 'in the news. Every time I see it, my reaction is probably: it's really impressive. And then? What does it have to do with me? Bitcoin to be finished?

These questions are not unanswerable, but rather most reports skip the most crucial steps: how does it actually work? Why fast? Where are you fast or not fast?

This article attempts to push from the most basic logic step by step to the current situation.

Let's start with the basics first. The core of a regular computer is "bits", a unit that can only be in a 0 or 1 state. Thousands of bits are arranged together and interact with each other through circuits, forming everything you can use: typing, watching videos, playing games.

The core of quantum computers is "quantum bits". The difference is that a quantum bit can be in a superposition state of 0 and 1 simultaneously.

'Overlay' is not a vague term like 'half 0, half 1', but strictly speaking, 'two possibilities exist simultaneously'. It's like when a coin is still spinning, you can't say whether it's on the front or the back, it's on top of the front and back. It wasn't until it landed, at the moment it was measured, that one side was randomly "selected".

The key here is: if you have one quantum bit, it can represent two states simultaneously; 2 quantum bits, 4 states; 3, 8 types; N qubits, 2 to the power of n states.

That is to say, for every additional quantum bit, the number of states you can handle simultaneously doubles. The number of states that can be processed simultaneously by 300 qubits is more than the total number of atoms in the universe. This is the source of all the power of quantum computing: exponential parallelism.

But there is an extremely important 'however' here: when you try to read the result from a quantum bit, the superposition state collapses into a definite value, 0 or 1. In other words, although you allow quantum bits to explore 2 ^ n possibilities simultaneously, you can only see one random result.

So the core challenge of quantum computing is to design algorithms that maximize the probability of the "correct result" appearing in the value you read. The entire history of quantum algorithms revolves around this problem.

Quantum computers are not faster than classical computers in all problems. In fact, in the vast majority of things, it is much slower than classical computers. It only has an overwhelming advantage in certain specific problems with strict mathematical proofs.

There are currently three recognized acceleration scenarios. The most commonly mentioned is large number decomposition, which is commonly referred to as "destroying encryption" by the outside world.

Give you two prime numbers, such as 17 and 19, and multiply them to get 323. You can calculate it in one second. But on the other hand, if we give you 323 and ask you to find out which two prime numbers are multiplied, it will be a bit troublesome. If you give a number that is not 323, but several hundred digits long, like the encryption used behind your bank card or WeChat Pay, the time required for a classical computer is similar to the age of the universe. It relies on the fact that 'decomposing large numbers is too difficult'.

In 1994, Peter Shor from MIT proposed a quantum algorithm that theoretically could decompose large numbers in just a few minutes. This is the origin of the statement that 'quantum computing will destroy existing encryption systems'. Note the wording: It destroys not the encryption itself, but the current mainstream RSA encryption system. Cryptologists are already preparing alternative solutions called 'post quantum cryptography', which will be mentioned later.

Another more practical and potentially earlier application is quantum simulation, which involves designing new materials and drugs. Why? Because molecules are essentially quantum mechanical systems. If you want to understand how a new drug molecule interacts with human proteins, using classical computer simulations is like using an abacus to calculate weather forecasts. It can be done, but it is extremely difficult and cannot be done on a large scale.

The quantum computer itself is a quantum system, and allowing it to simulate another quantum system is a natural fit for 'simulating the same class with the same class'. Specifically, you can "virtually" test the effectiveness of a new catalyst on a computer, without the need to test it thousands of times in the laboratory using bottles and jars. This is revolutionary for fields such as pharmaceuticals, materials science, and battery research and development.

Combination optimization is another direction that has been repeatedly discussed. The express delivery company faces a question every day: how can 200 cars go to 5000 points in the most fuel-efficient way? For each newly added point, the number of possible routes will double. The final 'solution space' is too large, and classical computers can only find a 'roughly good' solution.

Some specific quantum algorithms can theoretically find "better" solutions for optimization problems in certain structures. The optimization needs in logistics, financial investment portfolios, and chip design are trillion level, and even if only a few percentage points are optimized, the value is extremely huge.

First, let's give an intuitive positioning: current quantum computers are roughly equivalent to electronic tube computers in the 1940s, capable of running, verifying theories, and solving some specially designed problems, but still far from large-scale practical applications.

Google's progress is focused on error correction. Quantum bits are very fragile, and temperature fluctuations, electromagnetic interference, and even cosmic rays can make them go wrong. And the problem is that you put multiple qubits together to correct errors. In the past, the more you correct, the more error sources you bring.

At the end of 2024, Google's Willow chip, consisting of 105 qubits, was experimentally proven for the first time that as the number of qubits increased to a certain scale, the error rate began to exponentially decrease. This milestone of 'below the error correction threshold' has been pursued by physicists for nearly thirty years. This means that large-scale error correction is theoretically feasible.

However, it should be noted that Willow is demonstrating that "error correction itself is feasible", rather than "the quantum computer after error correction can already work". From principle verification to practical engineering, it is necessary to increase the number of physical bits by about three orders of magnitude.

The Ancestral Journey series of the Chinese team has taken another path and also succeeded. In 2025, the team led by Pan Jianwei from the University of Science and Technology of China released the "Zuchong III", which is also a superconducting processor with 105 qubits. On the benchmark test of random line sampling, the speed is 15 orders of magnitude faster than classical supercomputing. This result was published in the form of a cover paper in Physical Review Letters, and the reviewers evaluated it as "the highest level superconducting quantum computer currently available".

More importantly, by the end of 2025, they released "Zu Chong Zhi 3.2", which follows a different "full microwave control" route from Google and also achieves below the error correction threshold. This means that there are two independently validated technological paths in the world, which is a very healthy signal for the entire field.

IBM has chosen different routes. They do not pursue the highest performance of a single chip, but rather aim to enable the most people to use it. Currently, IBM has over 100 quantum computers deployed in the cloud, and anyone can run them by writing programs using Qiskit software. Their Condor chip has reached 1121 physical quantum bits, the highest number, with the goal of demonstrating "scientifically meaningful quantum advantages" by the end of 2026.

Microsoft's bet is the most aggressive. Everyone is using quantum error correction to deal with errors, but Microsoft wants to fundamentally eliminate them. What they do is' topological qubits', using a quasi particle called Majorana zero mode, which theoretically has the ability to resist errors. The Majorana 1 chip was released in early 2025, and a Nature paper at the same time reported key advances in single read measurements.

However, only 8 qubits have been verified so far, and it is still unknown whether they can be scaled up to practical scale. Moreover, the paper itself does not claim to have created a working topological qubit, which is highly controversial in the academic community. If successful, it may skip error correction and directly expand; If it doesn't work, it won't work forever. This is the highest risk bet in the entire race track.

The following may be the most important part of this article: what it cannot do.

It cannot replace your computer or phone. A quantum computer is not a 'stronger computer', it is a dedicated machine. For the vast majority of daily tasks, such as playing games, cutting videos, and browsing the web, it is much slower than your laptop. It won't make your phone faster or improve game graphics.

The current demonstration of 'quantum superiority' does not necessarily equate to practical ability. Whether it's Google's random line sampling or Zu Chongzhi III's benchmark testing, the core purpose of these tasks is to "prove that quantum computers do things that classical computers cannot do," but that "thing" itself has no practical use. This is a carefully designed proof question, not a practical application.

For example, being able to run 100 meters in 10 seconds proves that you are a world-class sprinter, but that doesn't mean you can lift bricks, deliver food, or fix pipes. From 'proving fast' to 'being able to do practical things', there are still several technical steps needed in between.

Fault tolerant quantum computers still require a long time. The current machines are all in the NISQ era, with the full name being Noisy Intermediate Scale Quantum, which refers to a medium scale quantum with noise. Easy to make mistakes, small in scale, and unable to operate stably for a long time.

Google and Zu Chongzhi's error correction breakthrough solves the problem of "from total errors to being able to correct errors". And from being able to correct errors to being able to do practical things, it is necessary to improve the efficiency of error correction by several orders of magnitude.

A specific estimate: To crack RSA encryption, approximately 20 million physical qubits are required. The current highest level of chips is over 1000. Even if error correction technology continues to advance, the industry generally believes that practical fault-tolerant quantum computers will still need 5 to 10 years.

What can it be used for? You can look at it in a timeline.

In the short term, from now until 2028, quantum chemistry simulations may be the first to be implemented. Hundreds of high-quality logical qubits can surpass classical computers in specific problems. The first step will be to help design better catalysts, which is related to the synthesis of ammonia and global food security; Understand the mechanism of high-temperature superconductivity; Accelerate drug molecule screening. You may not directly use it, but the drugs and materials you use in the future may indirectly come from its contribution.

In the medium term, portfolio optimization will begin to generate economic value from 2028 to 2033. With the maturity of error correction technology, quantum computers may find better solutions than classical algorithms in areas such as logistics path planning, supply chain optimization, and financial risk modeling. Not necessarily a 'flash sale classic', but a 'good 5% to 10%'. In an industry worth billions of dollars, this is already disruptive.

Looking further ahead, after 2033, the reconstruction of encryption systems and more fundamental scientific breakthroughs may occur. If a fault-tolerant machine at the level of millions of qubits is really created, mainstream encryption systems such as RSA need to be fully replaced. Many countries are already preparing for this matter, and NIST has released the post quantum cryptography standard FIPS 203/204/205 in 2024. Meanwhile, our understanding of the molecular world may enter an unprecedented depth.

But quantum computers are unlikely to be as popular as mobile phones. It is more like a supercomputer, built in a few national level laboratories and large enterprises, and remotely accessed through the cloud.

So, do you need anxiety now?

No need.

But if you look back five years later, you will find that the new materials born from quantum simulation, the carbon emissions saved by quantum optimization, and the new drugs created by quantum chemistry are the true traces left by quantum computing in history.

It did not come in a revolutionary way, but through infiltration.