Hello. Quantum computers are moving fast, aren’t they? In this article, I summarize the quantum computer industry.

4 hardware

Here are four types of hardware you need to know,

The first is superconducting, for which IBM and Google were competing. It is still widely used, and the price is much lower and easier to use than before.

The second is the ion trap, which is the method that is currently attracting the most attention. It is easy to use because of its low error, high coupling number, and long coherence time. The challenge will be how to improve the large scale and the long gate operation time.

The third is silicon qubits, the development of which is now in full swing. Although it is still at the stage of verifying the principle and creating a basic qubit, it is expected to overcome the weaknesses of superconducting qubits, such as the possibility of integration and operation at relatively high temperatures.

These three methods have the same calculation principle. They apply logical qubits to calculate. On the other hand, the fourth, optical quantum computer, has a slightly different computational principle, and its specifications have not yet been defined.

Each of these four is aiming to be an error-correcting quantum computer, making it the perfect environment in which to learn.

4 software

There are four steps to remember in software. Software can be divided into several stages of development.

The first is the stage where development has proceeded without hardware, assuming an ideal quantum computer. Basically, two major algorithms have been developed. These are quantum phase estimation, which is used in quantum chemistry and cryptanalysis, and quantum amplitude amplification, which is used in search. I have described three of them, plus a time evolution for quantum simulation.

In 2012, a quantum annealing machine specialized for combinatorial optimization problems was commercialized from Canada.

Around 2015, the race to develop a quantum computer between Google and IBM heated up, with the development of an algorithm using an error-prone machine called NISQ. On the NISQ machine, the originally conceived algorithm did not work due to an error, so an alternative algorithm was developed. VQE took the place of quantum phase estimation, and QAOA was a combination of quantum annealing and VQE that incorporated time evolution. Although quantum amplitude amplification is not very original for NISQ, it has been found to be useful for Monte Carlo calculations for finance, and is expected to remain an FTQC algorithm for future use. Then, quantum machine learning was born as a derivative of the quantum-classical hybrid algorithm developed from VQE, and a type of machine learning called Born machine is mainly flourishing. As of 2021, we should be able to remember four of these NISQ algorithms.

Quantum computers are already entering the next era, and since NISQ wasn’t that fast with hybrids, the companies’ hardware is already providing a roadmap for developing machines with fewer errors, called error correction. Correspondingly, some software is also shifting from hybrid to developing algorithms for FTQC.

4 cloud system

Quantum computers, which have so far only been used by researchers and a few companies, have been greatly developed as a cloud service after 2020.We will introduce the services of each company as they become available in 2021.

First, IBM has been focusing on quantum computing from the beginning and has already surrounded itself with many tutorials and users; IBM’s hallmark is vertical integration, doing everything from hardware to software, and other business development.

Amazon has changed the situation drastically from the vertically integrated business that has been the mainstream until now. Braket, released in 2020, changes the system of vertical integration that has been the mainstream until now to horizontal division of labor, allowing users to choose from several types of hardware. It leverages the strengths of aws, the world’s number one public cloud system.

Two more cloud systems are scheduled to launch this year, bringing the total to four.

Microsoft builds on Azure, the world’s second-largest public cloud, with a strong windows community and honeywell’s latest ion trap machine. Google will be the last service to integrate with GCP, the fourth largest public cloud service. Google integrated with tensorflow, a very powerful machine learning framework that will allow for the use of quantum computers in AI.

NISQ to FTQC

As the roadmaps of various companies are moving towards error correction, the development of applications will change rapidly, and the time when both NISQ and FTQC algorithms can be utilized and used in real quantum computers will be a dream come true for developers.