IBM’s Roadmap Targets a 4,000+ Qubit Quantum Computer
Up next on this year's agenda: a 433-qubit processor.
This week, IBM released a roadmap outlining the trajectory of its quantum computing development goals over the next few years.
Roadmap showing projected capabilities of IBM quantum computers over several years. Image courtesy of IBM
Central to this roadmap is IBM's focus on the superposition of qubits. A qubit is the quantum-mechanical analog of a classical bit. But unlike a classical bit (which can either take a value of 1 or 0, “on” or “off”), a qubit can take on a value that is a superposition of both states. By manipulating several qubits in superposition, quantum computers are able to perform a vast number of calculations much more quickly than a classical computer using the same number of bits.
Bloch sphere showing various states of a qubit represented as a 3D vector space. Image courtesy of Smite-Meister [CC BY-SA 3.0]
A 4,000+ Qubit Quantum Computer
By 2024, IBM aims to manufacture a 4,000+ qubit quantum computer called Kookaburra that will be accessible to developers via a cloud-based Qiskit Runtime. Qiskit is a quantum computing SDK that aims to abstract away the very complex details of quantum computing hardware and allow developers to focus on the task at hand.
Representation of a quantum circuit. Image courtesy of IBM
In particular, Qiskit allows developers to easily leverage a library of pre-built quantum circuits to perform complex calculations. A quantum circuit is a computational routine consisting of operations on qubits as well as classical computation. By providing Qiskit as a cloud-based service called Qiskit Runtime, IBM aims to allow developers to remotely run their routines on real quantum hardware from anywhere in the world.
Osprey and Condor Quantum Processors
While a 4,000+ qubit quantum computer is IBM’s stated goal by 2024, the company is working on two quantum processors for release in 2022 and 2023: Osprey and Condor.
The IBM Osprey is a 433-qubit quantum computer slated for release in the latter half of 2022. The processor is said to feature more efficient and denser controls and cryogenic infrastructure to increase the stability of the system and reduce noise. The IBM Condor is a 1,000+-qubit quantum computer that uses a large dilution refrigerator to cool the superconducting hardware of the quantum computer.
Layout showing a five-qubit device. Image courtesy of IBM
In IBM quantum computers, cooled superconductors are used to create qubits, and magnetic flux noise can be a major source of computational error. Therefore, it is essential to find ways to protect the quantum hardware from such sources of environmental noise.
Circuit Knitting for Running Large Circuits on Small Hardware
According to IBM, one of the major challenges of working with current quantum computers is figuring out how to run larger—say, 400-qubit quantum circuits—on 100-qubit systems. Circuit knitting is a technique that enables this by partitioning a large quantum circuit into smaller circuits capable of running on smaller-scale hardware.
Circuit knitting. The right image is a partitioned quantum circuit. Image courtesy of IBM
Such partitioning comes at the added cost of classical computation because each of the partitioned results is fed to a classical simulator to find the final result. While the cost of such knitting scales with the amount of partitioning required, it is still a viable technique to run larger qubit circuits until sufficient quantum hardware becomes available.
Next Steps Toward IBM's Quantum Computing
IBM initially announced its quantum roadmap in 2020—charting progress from the 27-qubit Hummingbird processor in 2019 to the 1,121-qubit Condor planned for 2023. IBM has already rolled out several of these processors on schedule, including last year's 127-qubit Eagle.
Expanded view of the Eagle processor. Image used courtesy of IBM
IBM has identified modular quantum computing as a key focus to reaching these ambitious hardware goals. Modular quantum computing entails that IBM:
- Find a way to classically communicate and parallelize operations across several processors
- Release short-range, chip-level couplers to create one larger processor
- Build quantum communication links between processors
With an eye on the 1,000+-qubit system, IBM also built a so-called super-fridge called "Goldeneye" in 2020, measuring 10-feet tall and 6-feet wide. This dilution refrigerator is larger than any commercial model today and will house a million qubits.
Featured image used courtesy of IBM.