What Is a General-Purpose Quantum Computer?

27 min readApr 26, 2023

There are a variety of types of quantum computers, but this informal paper focuses on defining what it means to be a general-purpose quantum computer, as opposed to a special-purpose quantum computing device. But not all general-purpose quantum computers will be practical quantum computers, although all practical quantum computers are indeed general-purpose quantum computers.

Topics discussed in this paper:

In a nutshell
What is a general-purpose quantum computer?
Glossary entry for general-purpose quantum computer
Quantum computing and simulation
Computation vs. simulation
Digital computing vs. analog computing
Two approaches to computation: analytical methods vs. numerical methods
Types of classical data: numbers, text, Booleans, and composite structures
Simulation on a classical computer vs. a quantum computer
Transistors and qubits: switching vs. amplification
Digital mode and analog mode for quantum computing
Analog mode using qubits: continuous value of probability amplitude
Exploiting the continuous analog values of probability amplitude and interference for quantum parallelism for computation
IBM’s utility-scale quantum computing focuses on simulation, not true computation
Quantum algorithms vs. quantum circuits
What are the rules of thumb for whether to use computation or simulation? The same as the rules for a knife fight. There are none!
Two modes of using a quantum computer: Computation vs. simulation
Quantum computer types
General-purpose quantum computers
Universal general-purpose gate-based quantum computer
Vendors of general-purpose quantum computers
Special-purpose quantum computing devices
Special-purpose quantum computers or special-purpose quantum computing devices?
General-purpose vs. special-purpose quantum computers
Don’t get confused by special-purpose quantum computing devices that promise much more than they actually can deliver
What’s wrong with analog computers?
Able to execute the full range of applications expected for quantum computing
Go for it if a special-purpose quantum computing device fits your needs, but that will likely be more of a rarity than common
Quantum computer as a coprocessor rather than a full-function computer
Quantum computers cannot fully replace classical computers
Universal quantum computer is an ambiguous term
Still not a universal quantum computer
Practical quantum computing isn’t really near, like within one, two, or three years
A practical quantum computer would be ready for production deployment to address production-scale practical real-world applications
All practical quantum computers are general-purpose quantum computers
Not all general-purpose quantum computers are practical quantum computers
NISQ devices can be general-purpose quantum computers
But the real future of general-purpose quantum computers is in post-NISQ quantum computing
24 to 40-qubit algorithms as the starting point for practical quantum computing
Quantum algorithms using less than 24 qubits simply don’t represent practical quantum computing
My own proposal for 48 fully-connected near-perfect qubits
Further information on quantum computers
Conclusions

In a nutshell

General purpose means that it applies to the widest range of applications.
And can be programmed for the most general quantum algorithms.
Quantum computing encompasses both computation, using either analytical methods or numerical methods, as well as analog mode simulation, but the focus here is exclusively on computation, not analog mode simulation.
In contrast to special-purpose quantum computers or special-purpose quantum computing devices. They apply to a more limited range of applications and have a much more limited ability to support arbitrary quantum algorithms. Includes quantum annealers, adiabatic quantum computers, boson sampling devices, photonic quantum computing, neutral-atom qubits, physics simulators, and software-configurable physics experiments.
Don’t get confused by special-purpose quantum computing devices that promise much more than they actually can deliver.
Able to execute the full range of applications expected for quantum computing.
What’s wrong with analog computers? Many or most of the special-purpose quantum computing devices are in fact analog computers, or operate in analog processing mode. Analog computers or analog processing mode can be great in some limited situations, but general-purpose computers can be great for a far wider range of applications.
Go for it if a special-purpose quantum computing device fits your needs, but that will likely be more of a rarity than common. But, be very careful — assessing application needs and matching to device capabilities can be very tricky. It’s an easier task to choose to go with a general-purpose quantum computer.
Synonym for universal gate-based quantum computer.
Still a coprocessor rather than a full-function computer.
Cannot fully replace a classical computer.
Still not a universal quantum computer. That’s a giant leap beyond a general-purpose quantum computer, which is still simply a coprocessor without all of the advanced capabilities of a general-purpose Turing machine, such as rich data types, control structures (loops, conditionals, nested function calls, and processes and inter-process communication), data structures, file systems, databases, network services, etc.
Practical quantum computing isn’t really near, like within one, two, or three years.
A practical quantum computer would be ready for production deployment to address production-scale practical real-world applications.
All practical quantum computers are general-purpose quantum computers. By definition.
Not all general-purpose quantum computers are practical quantum computers. Have the necessary functional capabilities, but insufficient capacity, such as too few qubits, too low qubit fidelity, too limited qubit connectivity, too limited coherence time, or fail to support reasonably large quantum algorithms.
NISQ devices can be general-purpose quantum computers. Technically, this is true, since general-purpose doesn’t strictly mean that it is a practical quantum computer.
The real future of general-purpose quantum computers is in post-NISQ quantum computing. In practical quantum computers. By definition, no longer NISQ devices.
24 to 40-qubit algorithms as the starting point for practical quantum computing. But a change in mindset will be required.
Quantum algorithms using less than 24 qubits simply don’t represent practical quantum computing. There might be some niche applications that can achieve some sort of quantum advantage and deliver some sort of business value using less than 24 qubits, but they would be the very rare exception rather than the norm.
My own proposal for 48 fully-connected near-perfect qubits. It may not be the most optimal configuration, but it’s a place to start. This is what I believe should be the low-end for a practical quantum computer. And would fairly represent general-purpose quantum computers.

What is a general-purpose quantum computer?

Cutting to the chase, with details to follow:

A general-purpose quantum computer is a quantum computing device which can be applied to the widest range of applications and programmed for the most general quantum algorithms.
This is in contrast with a special-purpose quantum computer or special-purpose quantum computing device which applies to a more limited range of applications and has a much more limited ability to support arbitrary quantum algorithms.
Universal gate-based quantum computer is a synonym for a general-purpose quantum computer.

Glossary entry for general-purpose quantum computer

This is a bit more wordy, but intended to capture more of the nuances:

general-purpose quantum computer. A quantum computer which has sufficient capabilities to usefully apply to a wide range of practical, real-world problems, in contrast to a fixed-function quantum computer which only applies to a narrow niche of problems. May simply be a synonym for large-scale quantum computer. It may or may not be a true universal quantum computer, able to compute whatever a classical computer can compute, but it does need to apply to a fairly wide range of problems even if it can’t replicate all functions of a classical computer. It will also need a large enough number of qubits, a reasonably long coherence, and some significant degree of quantum error correction. Even then, there may still be applications which can be executed on a classical computer but are still not able to execute on any existing general purpose quantum computer, such as those processing large volumes of semi-structured data or running complex software systems. See also: current general purpose quantum computer.

It and other glossary entries related to quantum computing can be found in my glossary:

Quantum computing and simulation

People frequently refer simply to quantum computing, but there are really two distinctive capabilities lumped together under the rubric of computing:

Computation. Generally, a computation consists of algebraic calculation — numeric calculation.
Simulation. Analog mode. No algebraic calculation per se.

In most contexts we don’t need to concern ourselves about the distinction, but sometimes it matters very much.

In any case, the concept of a general-purpose quantum computer espoused by this informal paper focuses exclusively on computation, not simulation.

Computation vs. simulation

To summarize the distinction:

Computation. Generally, a computation consists of algebraic calculation — numeric calculation. For numerical data, analytical methods or numeric methods are used to arrive at a numeric result, either exact or approximate. Or for non-numeric data, evaluation, manipulation, or generation of text, Booleans, or composite data — and numeric data too.
Simulation. Analog mode. Such as the evolution of Hamiltonians. No computation per se, in the algebraic sense.

It is indeed possible to do some simulations through computation, such as on a classical computer, but not all simulations are computations per se in the algebraic sense.

In any case, the concept of a general-purpose quantum computer espoused by this informal paper focuses exclusively on computation, not simulation.

Digital computing vs. analog computing

Sometimes the distinction between computation and simulation comes down to simply the distinction between:

Digital computing. Discrete values.
Analog computing. Continuous values.

Two approaches to computation: analytical methods vs. numerical methods

Classically, scientists and engineers have had two main approaches to calculation of numeric values:

Analytical methods. An exact answer can be calculated directly. Typically with an algebraic formula — fill in the numbers and directly calculate the result. Or, evaluate all possible solutions or using a heuristic such as a binary search, assuming that can be done quickly enough.
Numerical methods. An exact answer cannot typically be achieved directly, or at least not easily. An approximation is usually considered sufficient. Typically indirect means are required, such as brute force, exhaustive evaluation, statistical sampling, Monte Carlo Simulation, or variational methods. Typically too much time or resources would be required to evaluate all possible solutions.

If neither approach is feasible, simulation may be possible. In any case, the concept of a general-purpose quantum computer espoused by this informal paper focuses on situations whether analytical methods or numerical methods are feasible — an exclusive focus on computation, not simulation.

Types of classical data: numbers, text, Booleans, and composite structures

Although we generally characterize computation as numeric calculation, classical computers work with a variety of types of data:

Numbers. Integers and real numbers.
Text. Simple strings, modest descriptions, or entire stories or books.
Booleans. Distinguish between true and false. Or a simple binary 0 or 1.
Composite structures. AKA data structures. Arrays, lists, bit strings, tables, structures, trees and graphs, where the elements are numbers, text, booleans, or nested or linked composite structures. Complex media types, such as voice, music, videos, images, and multimedia content are simply composite structures, albeit very complex composite structures.

Even the most complex computations and applications ultimately distill down to manipulating data of these types.

Numeric data alone may be sufficient for traditional scientific and engineering applications, but expanding applications to the commercial, financial, and consumer domains quickly raises the needs for text, Booleans, and composite structures.

System and infrastructure software also immediately raise the need for non-numeric data, especially complex composite structures. This includes operating systems, system utilities, system services, compilers and other software development tools, monitoring and system accounting software, etc.

Granted, quantum computers do not (yet) support such a full range of types of data or processing, but the starting point for computation is still numbers, numeric calculation.

Simulation on a classical computer vs. a quantum computer

Simulation on a classical computer has no choice but to start with these same types of data, with a reliance on composite structures for much of the simulation, even if there is still a heavy reliance on numeric data for the actual physical phenomena being simulated or modeled.

In short, simulation on a classical computer is indeed a computation in the strictest sense.

But simulation on a quantum computer is rather different from computation on a classical computer.

I won’t go into the details here, but the essence of simulation on a quantum computer is a variant of analog computing, where each qubit is used to directly represent some physical quantity, such as an atom or a site for a chemical bond of a molecule, rather than a single bit of a larger item of numeric data. And the operations on these qubits are analog operations, comparable to the evolution of a Hamiltonian, rather than the operations typically performed on numeric, text, Boolean, or composite data on a classical computer.

Qubits used for simulation on a quantum computer are used in a manner which is much closer to the op-amps (operational amplifiers) used in a classical analog computer.

Transistors and qubits: switching vs. amplification

People familiar with computers are taught about only half of the function of transistors, namely that they are useful for switching, for conditionally and unconditionally transitioning between binary 0 and 1.

But transistors, similar to the vacuum tubes which preceded them, have two, distinct functions or capabilities:

Switching. Their binary, Boolean capability to represent 0 and 1, and transition between those two discrete states, sometimes unconditionally, but sometimes conditionally.
Amplification. Their ability to amplify a weak signal to a stronger signal, sometimes in a fixed manner, and sometimes a variable degree of amplification. Signals are continuous values, not the discrete 0 and 1 of binary switching.

Both capabilities can be used for computing:

Digital computing. Based on switching.
Analog computing. Based on amplification.

Qubits can be used in these two modes as well.

In fact, at the low level of the raw hardware, qubits function more like the transistors of a classical computer than the high-level bits of a classical computer.

Digital mode and analog mode for quantum computing

In fact, we see two modes of programming quantum computers:

Digital mode. Qubits act as classical bits, plus superposition, entanglement, and interference to exploit quantum effects and enable quantum parallelism, but all in this digital sense.
Analog mode. Qubits are used more like the op-amps of a classical analog computer, exploiting the amplification capability of the continuous values of probability amplitude to effect the evolution of Hamiltonians.

Some quantum computers are designed from the get-go to run in digital mode or analog mode, but not both.

Some quantum computers are designed to be able to run in either digital mode or analog mode — the user’s choice, as a configuration option.

And some quantum computers, such as superconducting transmon qubit machines, support both at the same time, and it’s simply a matter of how the algorithm uses quantum logic gates, either to exploit their binary switching capability or to exploit their continuous value probability amplitude to evolve Hamiltonians.

In any case, the concept of a general-purpose quantum computer espoused by this informal paper focuses exclusively on digital mode computation, not analog mode simulation.

Analog mode using qubits: continuous value of probability amplitude

Qubits are generally discrete in nature, focused on their strictly binary basis states, |0> and |1>.

But the quantum state of a qubit (or entangled qubits) also has probability amplitude, which is a continuous value, from 0.0 to 1.0, which effectively gives a qubit (or entangled qubits) an analog processing mode as well as it’s discrete, binary basis states.

Generally, probability amplitude is used in a manner which retains the discrete values of qubits. This is digital mode.

But some applications may choose to exploit this continuous value to effect an analog mode of processing, even when using the same exact qubits as are used for digital mode.

In any case, the concept of a general-purpose quantum computer espoused by this informal paper focuses exclusively on digital mode, not analog mode or analog computing.

Exploiting the continuous analog values of probability amplitude and interference for quantum parallelism for computation

Despite the obvious utility of the continuous values of probability amplitude for analog processing mode, it can also be used in a granular discrete mode to enable quantum parallelism for computation for digital mode.

Simply using the midpoint of probability amplitude, where the two basis states have equal (50%) probability, enables n fully entangled and superimposed qubits to represent 2^n discrete quantum states (n-bit bit string values), so that even a simple digital calculation can be performed on all 2^n values simultaneously.

It does require tricks and cleverness to select and retrieve a desired value from those 2^n simultaneous evaluations, such as using quantum Fourier transform, which further exploits finer gradations of the probability amplitude, as well as interference of phase. This enables quantum parallelism for computation.

So the analog processing mode of probability amplitude does have utility even for digital calculations on a quantum computer.

IBM’s utility-scale quantum computing focuses on simulation, not true computation

Recently, IBM has made a big deal about utility-scale quantum computing with their 100-qubit quantum computers (127-qubit Eagle and 133-qubit Heron), but it turns out that the 100-qubit circuits they are promoting in their recent papers are really focused on simulation in the analog sense, not true computation in the algebraic, analytical sense that we have just discussed here.

Indeed, there can be significant value to analog mode simulation, but it’s not representative of general-purpose computation as espoused here.

The IBM quantum computers are in fact general-purpose quantum computers, but they don’t support general-purpose computations at the 100-qubit scale.

In fact, the Quantum Volume (QV) for Eagle and Heron are 128 and 512, respectively, indicating that they work well in the general-purpose, algebraic, sense only up to roughly 7 and 9 qubits, respectively. So, general purpose, but still in a very limited sense.

Quantum algorithms vs. quantum circuits

Whether a quantum computer is being used for general-purpose computation or analog-style simulation, it will still use a quantum circuit.

But just because a quantum circuit is being used does not mean that general-purpose computation is being performed.

Generally, algorithms are associated with general-purpose computation. An algorithm implies a computation, and computation implies an algorithm.

But a circuit does not necessarily imply an algorithm — or a computation.

So, in the case of IBM and various papers, a 100-qubit circuit used for simulation does not mean that they are performing a 100-qubit algorithm or computation.

What are the rules of thumb for whether to use computation or simulation? The same as the rules for a knife fight. There are none!

I would expect that as quantum computing slowly matures, rules of thumb and decision flow charts will evolve to make it easier to decide whether a particular application problem is more suited for classical computing, analytical methods, numerical methods, or classical or quantum simulation.

But we’re definitely not there, yet.

Right now, quantum computing and algorithm design are suited only for the most senior and elite of technologists.

Much trial and error, intuition, judgment, and technical cleverness is required.

Two modes of using a quantum computer: Computation vs. simulation

With all of that background, now we make the main point that it is important to distinguish these two modes of using a quantum computer:

Computation. For numerical data, analytical methods or numeric methods to arrive at a numeric result, either exact or approximate. Or for non-numeric data, evaluation, manipulation, or generation of text, Booleans, or composite data — and numeric data too.
Simulation. Analog mode. Evolution of Hamiltonians. No computation per se.

In any case, the concept of a general-purpose quantum computer espoused by this informal paper focuses exclusively on computation, not analog mode simulation.

Quantum computer types

There are really two very distinct categories of quantum computers:

General-purpose quantum computer. Can be applied to many different types of applications. May be referred to more technically as a universal general-purpose gate-based quantum computer.
Special-purpose quantum computing device. Only applies to a relatively narrow niche of applications which meet criteria of the device. Some may refer to it as a special-purpose quantum computer as well, although its lack of generality argues against this. May also be referred to as a single-function quantum computing device or a single-function quantum computer.

For the most part, this paper focuses on the former, general-purpose quantum computer.

General-purpose quantum computers

General-purpose quantum computers are the most general. They are the most flexible. And generally the most useful.

A general-purpose quantum computer may also be more technically specifically described as a universal general-purpose gate-based quantum computer. The extra adjectives clarify the distinctions from special-purpose quantum computing devices.

For the purposes of this paper we are focused almost exclusively on general-purpose quantum computers.

Universal general-purpose gate-based quantum computer

Just to emphasize that universal general-purpose gate-based quantum computer is a technically specific term which distinguishes general-purpose quantum computers from special-purpose quantum computing devices (or special-purpose quantum computers).

Vendors of general-purpose quantum computers

Some of the vendors of general-purpose quantum computers include:

IBM.
Rigetti Computing.
IonQ.
Honeywell.
Google.
Intel.

That’s not intended to be an all-inclusive list, and focuses too heavily on American companies, just some examples to help the reader navigate the field.

Special-purpose quantum computing devices

Some of the different types of special-purpose quantum computing devices (or special-purpose quantum computers) that exist today:

Quantum annealers. Focused on quantum annealing (QA). Speciality is a niche of optimization problems. Not a general-purpose quantum computer for non-optimization applications. D-Wave Systems is the best example.
Adiabatic quantum computers. Generalization of quantum annealers. Again, not general-purpose.
Boson sampling devices. Another specialized form of quantum computing — a special-purpose quantum computer or special-purpose quantum computing device. Also known as Gaussian boson sampling (GBS).
Photonic quantum computing. Sometimes simply a synonym for boson sampling device. Still too early in the research stage to judge whether it has potential as a general-purpose quantum computer or is restricted to special purposes such as boson sampling devices.
Neutral-atom qubits. QuEra, Cold Quanta (Infleqtion), Atom Computing, Pasqal. Maybe, theoretically they could eventually be used for truly general-purpose quantum computing in much the same way as trapped-ion qubits, but at present they tend to be used in more of an analog processing mode or more as software-configurable physics experiments rather than true general-purpose computing.
Physics simulators. Custom-designed, special-purpose quantum computing devices designed to address specific physics problems. Not general-purpose quantum computers. Debatable whether they should even be called quantum computers, but they are specialized quantum computing devices.
Software-configurable physics experiments. A step up from task-oriented physics simulators, these physics experiments are programmable or at least configurable to some degree, but simply in terms of varying the physics experiment, not intending to offer a general-purpose quantum computing capability as is the case with gate-based superconducting transmon qubits or trapped-ion qubits. Current neutral-atom quantum computers, typically running in so-called analog processing mode, are effectively software-configurable physics experiments.

Special-purpose quantum computers or special-purpose quantum computing devices?

Whether to refer to special-purpose quantum computing devices as special-purpose quantum computers is an interesting semantic challenge and debate. There’s no clear, bright-line answer at this stage.

Some will prefer the latter and even simply refer to them as quantum computers despite their lack of being general purpose.

I will default to overall referring to them as special-purpose quantum computing devices, but I won’t object too strenuously if people refer to them as special-purpose quantum computers. I will object if people refer to them simply as quantum computers.

General-purpose vs. special-purpose quantum computers

Just to emphasize the point that when someone, especially in the media, but even in technical media, uses the term quantum computer that doesn’t immediately tell you whether they may be talking about a general-purpose quantum computer or a special-purpose quantum computing device (or special-purpose quantum computer) as distinguished in the preceding sections.

Don’t get confused by special-purpose quantum computing devices that promise much more than they actually can deliver

Special-purpose quantum computing devices can indeed deliver amazing capabilities, but only for a relatively narrow niche of functions. If your particular application problem can’t be easily mapped to that narrow niche, you’re out of luck. You’re probably better off with a general-purpose quantum computer, in general.

The catch is that we’re still very early in the development and evolution of quantum computers, so that even general-purpose quantum computers aren’t yet up to solving many application problems.

What’s wrong with analog computers?

There’s nothing wrong with analog computers per se, but… it all depends on a lot of factors.

Analog computers have been around for as long as if not longer than classical digital computers. In some cases they work quite well. In other cases, or even the general case, not so much.

Many or most of the special-purpose quantum computing devices are in fact analog computers, or operate in analog processing mode, such as described in a previous section:

Quantum annealers.
Adiabatic quantum computers.
Boson sampling devices.
Photonic quantum computing.
Neutral-atom qubits.
Physics simulators.
Software-configurable physics experiments.

The upsides of analog computers:

They can rapidly deliver useful answers, sometimes even faster than a classical digital computer.
They can be just what’s needed when only an approximate answer is needed.
They can be relatively simple.
They can be relatively cheap.
They can be feasible even when a more general-purpose computer is not yet as feasible.

The downsides:

They deliver only approximate solutions.
They lack fine precision in both input data, intermediate results, and final results.
They lack very small and very large magnitudes for both input data, intermediate results, and final results. A limited range of magnitude.
They are far from ideal.
They lack generality.
They tend to be focused on niche problems or offer only niche solutions.
They can be very complex.
They can be very difficult to use if your problem isn’t a very close match for their capabilities.
They can be very expensive.
They may not be able to deliver a solution in many of the cases when a general-purpose computer can.

To recap a few of the most critical downsides:

Only approximate solutions.
Lack of fine precision.
Lack of a wide range of magnitude. A limited range of magnitude.
Lack of generality. For specialized problems.

In short, sure, analog computers or analog processing mode can be great in some limited situations, but general-purpose computers can be great for a far wider range of applications.

Able to execute the full range of applications expected for quantum computing

Unlike special-purpose quantum computing devices which can execute only a limited range of applications, a general-purpose quantum computer is able to execute the full range of applications that have been promised for quantum computing.

Although this may be subject to the caveat of not necessarily supporting some applications that may require more qubits, greater qubit fidelity, greater qubit connectivity, or greater maximum circuit size than the device supports.

For a summary of the types of applications that have been promised for quantum computing, see my informal paper:

Go for it if a special-purpose quantum computing device fits your needs, but that will likely be more of a rarity than common

There are indeed going to be situations where a special-purpose quantum computing device may fit your needs, but that will likely be more of a rarity than a common situation.

Be very careful — assessing your application needs and matching them to the capabilities of a specific special-purpose quantum computing device can be extremely tricky. A task best left to the most elite of experts, not the average quantum algorithm designer or quantum application developer.

It’s a much easier task to choose to go with a general-purpose quantum computer.

If you fail with a general-purpose quantum computer, you can easily be forgiven, but if you fail by rejectinge a general-purpose quantum computer and going with a special-purpose quantum computing device, people will be more likely to ask what were you thinking.

A general-purpose quantum computer is the safer choice.

Quantum computer as a coprocessor rather than a full-function computer

We commonly refer to a quantum computer as if it was a full-function computer system, but it falls far short of that. There are many functions which can be performed with even the simplest classical computer which are beyond what is possible with even the best quantum computers. But the whole point of a quantum computer or quantum processor if you will is that it performs only a small fraction of what a classical computer can do, but it performs that fraction extremely well, even far better than even the best classical computers.

Most of the function of a typical quantum application is performed on a classical processor, and only a small fraction of its function is offloaded to a quantum processor.

The quantum processor acts as an auxiliary processor.

This is similar to the way that graphical functions can be offloaded to a graphical processing unit (GPU).

The term coprocessor is a synonym for auxiliary processor.

So the bulk of the quantum application is processed by the classical computer as the main processor with a small fraction of the function offloaded to the quantum processor (quantum computer) as a coprocessor.

In short, what we commonly refer to as a quantum computer or quantum processor or quantum processing unit or QPU is in fact an auxiliary processor or coprocessor.

Quantum computers cannot fully replace classical computers

Quantum computers may sound as if they have great potential to fully replace classical computers, but that just isn’t true. Although quantum computers can do some tasks much better than classical computers, there are many tasks that classical computers can do that quantum computers simply can’t do at all.

Some examples of processing tasks which are not appropriate for quantum computers but are appropriate for classical computers, see my informal paper:

Eventually quantum computers will be merged with classical computers as a universal quantum computer, but that’s far in the future, at least not real soon.

For my own proposal for a universal quantum computer, see my informal paper:

Universal quantum computer is an ambiguous term

Saying that a quantum computer is universal is ambiguous. There are two distinct meanings:

Universal gate set. The machine is capable of all possible functional manipulations of quantum information. Implies general purpose, in contrast to special purpose. Implies a general-purpose quantum computer. This fits for most current gate-based quantum computers. It’s now a standard feature. An expected feature.
Universal quantum computer. A proposal for a merger of quantum computing and classical computing into a single integrated machine. Not two separate machines interfacing, but fully integrated into a single machine with classical and quantum data coexisting.

It’s not uncommon for people, including published researchers, to refer to any quantum computer with a universal gate set as a universal quantum computer. This is a misuse of the term, but unfortunately it is fairly common. They should simply use the term general-purpose quantum computer.

For my own proposal for a universal quantum computer, see my informal paper:

Still not a universal quantum computer

A universal quantum computer is a proposal for a merger of quantum computing and classical computing into a single integrated machine. Not two separate machines interfacing, but fully integrated into a single machine with classical and quantum data coexisting.

That’s a giant leap beyond a general-purpose quantum computer, which is still simply a coprocessor without all of the advanced capabilities of a general-purpose Turing machine, such as rich data types, control structures (loops, conditionals, nested function calls, and processes and inter-process communication), data structures, file systems, databases, network services, etc.

Practical quantum computing isn’t really near, like within one, two, or three years

Although quantum computers do exist today, they are not ready for production deployment to address production-scale practical real-world applications. And it’s unlikely that they will be in the next one, two, or three years.

Sure, there may be some smaller niches where they can actually be used productively, but those would be the exception rather than the rule.

Four to seven years is a better bet, and even then only for moderate benefits.

Generation of true random numbers is one exception where quantum computers are actually commercially viable today, although technically a full-blown quantum computer is not needed just to use quantum effects to generate random numbers.

A practical quantum computer would be ready for production deployment to address production-scale practical real-world applications

Just to clarify the terminology — we do have quantum computers today, but they haven’t achieved the status of being practical quantum computers since they are not yet ready for production deployment to address production-scale practical real-world applications.

We’re unlikely to see a practical quantum computer in the next few years.

For more detail on practical quantum computers, see my informal paper:

All practical quantum computers are general-purpose quantum computers

By definition, all practical quantum computers are general-purpose quantum computers.

Not all general-purpose quantum computers are practical quantum computers

Although, by definition, all practical quantum computers are general-purpose quantum computers, not all general-purpose quantum computers are practical quantum computers due to insufficient capacity.

A general-purpose quantum computer might not be a practical quantum computer if either:

It has too few qubits. Insufficient to address production-scale practical real-world problems.
It has too low qubit fidelity. Ditto.
It has too limited qubit connectivity. Ditto.
It has too limited granularity of phase and probability amplitude for its qubits. Ditto.
It has too limited coherence time. Ditto.
It fails to support reasonably large quantum algorithms. Ditto.

NISQ devices can be general-purpose quantum computers

Technically, this is true, that NISQ devices can be general-purpose quantum computers since general-purpose doesn’t strictly mean that it is a practical quantum computer.

The ultimate goal is practical quantum computing, but a general-purpose quantum computer is but part of the path to get there.

But the real future of general-purpose quantum computers is in post-NISQ quantum computing

The real goal is not simply general-purpose quantum computing, but practical quantum computing, which generally implies or requires advancing beyond the limitations of NISQ devices, including qubit fidelity, qubit connectivity, granularity of qubit phase and probability amplitude, and maximum circuit size.

Post-NISQ quantum computing should meet the requirements for practical quantum computing. This is the environment in which quantum computing in general and general-purpose quantum computing in particular will thrive.

For more on post-NISQ quantum computing, see my informal paper:

24 to 40-qubit algorithms as the starting point for practical quantum computing

At present, few quantum algorithms focused on practical real-world applications use more than even 16 qubits. In fact, I haven’t seen any use even 24 qubits. Oddball computer science experiments excepted.

I would suggest that quantum algorithms using between 24 and 40 qubits are likely to be the low end and starting point for practical quantum computing.

Despite the fact that we can currently simulate quantum algorithms using 24 to 40 qubits, we still are not seeing any quantum algorithms in that range.

It will likely take a significant change in mindset before we do start seeing quantum algorithms in that range.

Quantum algorithms using 48 to 80 qubits are likely to enable even more practical quantum applications, but unfortunately we won’t be able to fully simulate them, which makes their development, debugging, and testing significantly more challenging.

Once again, it will likely take yet another significant change in mindset before we do start seeing quantum algorithms in that range.

And quantum algorithms using 80 to 100 qubits, or 100 to 120 qubits, or 120 to 160 qubits, or even 160 to 250 qubits are well beyond what we can speculate let alone actually support in the next few years.

For more on the dearth of practical quantum algorithms, see my informal paper:

Quantum algorithms using less than 24 qubits simply don’t represent practical quantum computing

While it’s credible to imagine that quantum algorithms using 24 to 40 qubits can represent practical quantum computing, at least as a low end starting point, it simply isn’t credible to believe that quantum algorithms using less than 24 qubits could be solving practical real-word problems and achieving a significant quantum advantage and delivering significant business value.

Some of the qubit capacities unlikely to represent practical quantum computing:

5 qubits.
8 qubits. Or 7 qubits.
10 qubits. Or 9 qubits.
12 qubits. Or 11 or 13 qubits.
16 qubits. Or 15 or 17 qubits.
20 qubits. Or 21 qubits.
22 qubits. Or 23 qubits.

There might be some niche applications that can achieve some sort of quantum advantage and deliver some sort of business value using less than 24 qubits, but they would be the very rare exception rather than the norm.

My own proposal for 48 fully-connected near-perfect qubits

The overall definition for a practical quantum computer covers a wide range and is relatively agnostic as to specific targets for the many essential factors.

I did endeavor to write up a proposal for what I believe should be the low-end for a practical quantum computer:

It may not be the most optimal configuration, but it’s a place to start.

I also speculated about a 36-qubit configuration as a stepping stone.

I also speculated about upgrading existing 27-qubit designs to this proposal as an even smaller initial stepping stone.

And I also speculated about a larger, 72-qubit configuration.

Further information on quantum computers

For further details on quantum computers, see my informal paper:

Conclusions

General purpose means that it applies to the widest range of applications.
And can be programmed for the most general quantum algorithms.
Quantum computing encompasses both computation, using either analytical methods or numerical methods, as well as analog mode simulation, but the focus here is exclusively on computation, not analog mode simulation.
In contrast to special-purpose quantum computers or special-purpose quantum computing devices. They apply to a more limited range of applications and have a much more limited ability to support arbitrary quantum algorithms. Includes quantum annealers, adiabatic quantum computers, boson sampling devices, photonic quantum computing, neutral-atom qubits, physics simulators, and software-configurable physics experiments.
Don’t get confused by special-purpose quantum computing devices that promise much more than they actually can deliver.
Able to execute the full range of applications expected for quantum computing.
What’s wrong with analog computers? Many or most of the special-purpose quantum computing devices are in fact analog computers, or operate in analog processing mode. Analog computers or analog processing mode can be great in some limited situations, but general-purpose computers can be great for a far wider range of applications.
Go for it if a special-purpose quantum computing device fits your needs, but that will likely be more of a rarity than common. But, be very careful — assessing application needs and matching to device capabilities can be very tricky. It’s an easier task to choose to go with a general-purpose quantum computer.
Synonym for universal gate-based quantum computer.
Still a coprocessor rather than a full-function computer.
Cannot fully replace a classical computer.
Still not a universal quantum computer. That’s a giant leap beyond a general-purpose quantum computer, which is still simply a coprocessor without all of the advanced capabilities of a general-purpose Turing machine, such as rich data types, control structures (loops, conditionals, nested function calls, and processes and inter-process communication), data structures, file systems, databases, network services, etc.
Practical quantum computing isn’t really near, like within one, two, or three years.
A practical quantum computer would be ready for production deployment to address production-scale practical real-world applications.
All practical quantum computers are general-purpose quantum computers. By definition.
Not all general-purpose quantum computers are practical quantum computers. Have the necessary functional capabilities, but insufficient capacity, such as too few qubits, too low qubit fidelity, too limited qubit connectivity, too limited coherence time, or fail to support reasonably large quantum algorithms.
NISQ devices can be general-purpose quantum computers. Technically, this is true, since general-purpose doesn’t strictly mean that it is a practical quantum computer.
The real future of general-purpose quantum computers is in post-NISQ quantum computing. In practical quantum computers. By definition, no longer NISQ devices.
24 to 40-qubit algorithms as the starting point for practical quantum computing. But a change in mindset will be required.
Quantum algorithms using less than 24 qubits simply don’t represent practical quantum computing. There might be some niche applications that can achieve some sort of quantum advantage and deliver some sort of business value using less than 24 qubits, but they would be the very rare exception rather than the norm.
My own proposal for 48 fully-connected near-perfect qubits. It may not be the most optimal configuration, but it’s a place to start. This is what I believe should be the low-end for a practical quantum computer. And would fairly represent general-purpose quantum computers.

For more of my writing: List of My Papers on Quantum Computing.