What is a QPU (quantum processing unit)?

As they contain the quantum part of quantum computers, QPUs can be used to help solve challenging problems facing humanity with the potential to impact climate change, pharmaceutical development and artificial intelligence (AI).

In the same way a central processing unit (CPU) can be thought of as “the brain of the computer” in classical computing, the quantum processing unit functions like “the brain” of quantum computing systems. Just as a CPU is more than just a chip and includes several other components, a QPU contains physical computational qubits as well as the control electronics and classical compute hardware used for holding instructions in memory, amplifying and managing input and output signals and separating signals from noise. 

The QPU is the core component of any quantum computer, and the quantum chip is the core component of a QPU. At IBM, the quantum chip is a multilayer semiconductor etched with superconducting components. These components are the physical qubits used to perform quantum calculations. These chips are further divided into multiple layers featuring the qubits, readout resonators and multiple layers of wiring for inputs and outputs. 

QPUs contain a quantum chip similar in size to your average computer chip—also known as a quantum data plane—composed of physical qubits laid out in various configurations, and the structures to hold them in place. The chip is held at cold temperatures near absolute zero in a dilution refrigerator.

QPUs also include control electronics and classical compute hardware required for input and output. Some of these components sit inside the dilution refrigerators, while other components sit in a rack at room temperature beside the dilution refrigerator.

QPUs are unique among computer processor units. Unlike CPUs, quantum processors take advantage of quantum physics to store and process data differently. Classical CPUs use binary bits to store data as either a 0 or a 1.

Qubits can store binary information in zeros and ones, but they can also hold a superposition, meaning they store a special combination of both 0 and 1. QPUs also take advantage of several other key quantum principles that allow them to process information in ways classical computers struggle to replicate.

Representing a generational advancement in computer science, QPUs are designed to process quantum algorithms better than even the most powerful supercomputers. Optimized for large-scale quantum calculations, QPUs are not intended to replace CPUs. Instead, QPUs are being integrated into high-performance computing (HPC) systems alongside CPUs and graphics processing units (GPUs). 

In a quantum-centric supercomputer, each type of processor functions differently and is used for processing different types of computations with the ecosystem:

• CPUs: Central processing units (CPUs) process inputs sequentially, performing tasks in a linear fashion and are best suited for high-level control operations, such as managing data across different system components.

• GPUs: Graphics processing units (GPUs) excel at parallel processing large amounts of operations simultaneously. GPUs can potentially be used in quantum systems to offload some amount of the processing workload from QPUs.

• NPUs: Neural processing units (NPUs) mimic the function of the human brain and are typically used for complex tasks involving the specific types of computations used in machine learning (ML) and artificial intelligence (AI).

• QPUs: Quantum processing units (QPUs) process information by using qubits instead of binary bits and are designed to perform complex quantum algorithms. QPUs are best used for certain kinds of highly complicated problems, and many of today’s promising quantum algorithms provide probabilistic solutions instead of precise answers.

Considered only theoretical in the 20th century, recent advancements in quantum technologies have led to a surge in QPU development. Today, IBM is pushing the boundaries of computer science to develop viable QPUs capable of achieving quantum advantage—the ability to outperform all classical supercomputing methods for solving a given problem. Developers at IBM are leading the pack, already delivering QPUs and quantum hardware with quantum utility—the ability to provide reliable, accurate outputs to quantum circuits beyond the reach of brute-force classical simulations. 

Quantum computing is an emergent technology that harnesses the power of quantum mechanics to solve problems too complex for even the most powerful supercomputers. Tasks like factoring large prime numbers, which might take a classical computer hundreds of thousands of years, can theoretically be accomplished in a matter of minutes with a sufficiently powerful quantum computer. 

Quantum computers process information differently than classical computers. Unlike classical computers that must compute every step of a complicated calculation with the rules of logic, quantum circuits made from qubits can process many entries of dataset simultaneously with quantum operations, providing a new way to tackle certain problems and potentially improving efficiency by many orders of magnitude. 

Classical computing: 

• Used by common, multipurpose computers and devices
• Stores information in bits with a discrete number of possible states, 0 or 1
• Processes data logically and sequentially

Quantum computing:

• Used by specialized and experimental quantum mechanics-based quantum hardware
• Stores information in qubits as 0, 1 or a superposition of 0 and 1
• Processes data with quantum logic at parallel instances, relying on entanglement and interference

Quantum processors do not perform mathematical equations the same way classical computers do. Unlike classical computers that must compute every step of a complicated calculation with the rules of logic, quantum circuits made from qubits process many entries of dataset simultaneously with quantum operations, providing a new way to tackle certain problems and potentially improving efficiency by many orders of magnitude.

Whereas traditional computers use transistors to store and process data in binary code, QPUs use qubits. IBM QPUs use solid-state superconducting qubits to encode data as either 0, 1, or a superposition of 0 and 1. As the number of qubits increases, every possible combination of all the qubits’ values can also be held in a superposition. Within these positions, certain qubits might become entangled, in which case their values become dependent on others, and they can no longer be considered to be behaving independently. Measuring one entangled qubit instantly provides information on the state of the other. Entanglement is a valuable tool for running quantum algorithms.

At the end of a quantum calculation, the data is converted by the QPU and supporting hardware into binary, and either a 0 or 1 will be measured on each qubit with a probability corresponding to its contribution to the superposition. 

Quantum technologies can use actual particles known as molecular qubits or hardware mimicking the behavior of particles (such as superconducting qubits) to perform calculations in ways that binary bits can’t, enabled by four key principles only found in quantum systems. 

1. Superposition: Superposition is the state in which a quantum particle or system can represent not just one value, but a combination of multiple possible values. 
2. Entanglement: Entanglement is the process in which two quantum particles become correlated more strongly than regular probability allows.
3. Decoherence: Decoherence is the process in which quantum particles and systems can decay, collapse or change, converting into single states measurable by classical physics.  
4. Interference: Interference is the phenomenon in which entangled quantum states can interact and produce more and less likely probabilities.

Generally, qubits are created by manipulating and measuring quantum particles (smallest known building blocks of the physical universe, such as photons, electrons, trapped ions and atoms) or by engineering systems that mimic these particles.

• Superconducting qubits: Made from superconducting materials operating at low temperatures, these qubits are favored for their speed in performing computations and fine-tuned control. 

• Trapped ion qubits: Trapped ions can also be used as qubits and are noted for long coherence times and high-fidelity measurements. Ions are atoms with electrical charge.

• Quantum dots: Quantum dots are small semiconductors that capture a single electron and use it as a qubit, offering promising potential for scalability and compatibility with existing semiconductor technology. 

• Photons: Photons are individual light particles used to send quantum information across long distances through optical fiber cables and are being used in quantum communication and quantum cryptography. 

• Neutral atoms: Commonly occurring neutral atoms charged with lasers are well suited for scaling and performing operations.

Certain types of qubits are better suited for certain tasks, although all known qubits are still highly sensitive. QPUs used in functional quantum computers require significant support hardware and software to maintain proper calibration and handle external noise. Software solutions like IBM’s Qiskit software stack feature tools that are used to orchestrate across quantum and classical hardware and perform necessary quantum error handling to help eliminate inaccurate readouts through automation.

While the chip inside of a QPU is about the same size as the chips in a typical CPU or GPU, quantum computing systems might be as large as a four-door sedan. This extra bulk mostly comes from cryogenic systems and refrigerators that must cool the qubits to temperatures colder than outer space to maintain coherence. It also includes other classical components used to send and apply instructions and return outputs, which can be stored at room temperature.  

Quantum computers powered by QPUs excel at solving certain complex problems, with the potential to speed up the processing of large-scale data sets. From the development of new drugs and performing machine learning (ML) in a new way to supply-chain optimization and performing time-series modeling on complex climate data, quantum computing might hold the key to breakthroughs in many critical industries.

QPUs will also be used in quantum-centric supercomputing to solve the most complicated and challenging problems facing humanity today in the following fields: 

• Pharmaceuticals: Quantum computers capable of simulating molecular behavior and biochemical reactions can massively speed up the research and development of new, life-saving drugs and medical treatments. 

• Chemistry: For the same reasons quantum computers can impact medical research, they might also provide undiscovered solutions for mitigating dangerous or destructive chemical byproducts. Quantum computing can lead to improved catalysts that enable petrochemical alternatives or better processes for the carbon breakdown necessary for combating climate-threatening emissions. 

• Artificial intelligence and machine learning: As interest and investment in AI and related fields like machine learning ramps up, researchers are pushing AI models to new extremes, testing the limits of our existing hardware and demanding tremendous energy consumption. There is evidence that some quantum algorithms might be able to look at datasets in a new way, providing a speedup for some machine learning problems.

• Materials science: Many materials science problems are innately quantum, and quantum speedups in this field have the potential to benefit areas from our fundamental understanding of matter to industrial problems in energy storage, solar power and more.

• Optimization: Efficient resource optimization offers value to any given industry; however, as logistics grow increasingly complicated, optimization becomes even harder. Quantum computers do not explore every solution in parallel—at least, not in the way that's useful for optimization. But that doesn’t mean that they can’t provide new solutions that are better than existing models. New research is emerging that demonstrates how and where quantum might provide value for optimization, and on what timeline. In fact, we already know that some quantum approximation algorithms run efficiently—in polynomial time—that give a solution that's 80% optimal.