The Great Quantum Computing Race

Companies and countries are pouring tens of billions of dollars into different qubit technologies, but it’s still too early to predict a winner.


Quantum computing is heating up, as a growing number of entities race to benchmark, stabilize, and ultimately commercialize this technology.

As of July 2021, a group from China appears to have taken the lead in terms of raw performance, but Google, IBM, Intel and other quantum computer developers aren’t far behind. All of that could change overnight, though. At this point, it’s too early to declare a winner in quantum computing, a technology that promises to outperform today’s conventional supercomputers.

Today, Google, IBM and others have built the first wave of quantum computers, but these systems are still in the early stages and aren’t yet running any useful commercial applications — yet. Nevertheless, there is noticeable progress with quantum computing, which is different than today’s systems.

In today’s computing, the information is stored in bits, which can be either a “0” or “1”. In quantum computing, the information is stored in quantum bits, or qubits, which can exist as a “0” or “1” or a combination of both. The superposition state enables a quantum computer to perform multiple calculations at once, enabling it to outperform a traditional system. But the technology faces a number of challenges, and many industry experts believe these systems are still a decade away from being practical.

However, that’s not stopping companies, governments, R&D organizations and universities from developing the technology and pouring billions of dollars into the arena. If they are realized, quantum computers could accelerate the development of new chemistries, drugs and materials. The systems also could crack any encryption, which has made their development a top priority among several nations. And across the board, it could provide companies and countries with a competitive edge.

“Quantum computing is at the forefront of national initiatives,” said Amy Leong, senior vice president at FormFactor. “There have been more than $20 billion in investments announced across 15 countries here. Geopolitical powerhouses like the U.S. and China are certainly leading the race to claim quantum supremacy, followed by a host of others from Europe and Asia.”

The race is heating up among nations as well as between different organizations. In a major development, the University of Science and Technology of China (USTC) in June 2021 demonstrated what researchers claim is the world’s fastest quantum computing processor, surpassing the previous and unofficial record held by Google’s 53-qubit device since 2019. USTC’s 66-qubit processor performed a complex calculation in 1.2 hours that would have taken today’s supercomputers 8 years to complete.

Google, IBM, Intel and other quantum computing developers aren’t standing still, and are aggressively devising faster processors. It’s too early to declare a winner, as the technology is still in its infancy. “When I take a look at the first applications, we’re going to need several thousand, if not 100,000 qubits, to do something useful,” said James Clarke, director of quantum hardware at Intel. “If we’re at 50 to 60 qubits today, it’s going to be a while before we can get to 100,000 qubits. It’s going to be awhile before we can get to 1 million qubits, which would be necessary for cryptography.”

Meanwhile, there is another race within this race. Vendors are developing a dozen types of qubits based on a range of technologies, such as ion trap, silicon spin and superconductivity. Vendors from each camp claim their technology is superior, and will enable practical quantum computers. It’s too early to declare a technology winner here, as well.

Still, the market is promising. The quantum computer market is projected to grow from $320 million in 2020 to $830 million by 2024, according to Hyperion Research.

Classical vs. quantum computing
Viewed as a timeline, the computing field has made enormous progress. In 1945, the University of Pennsylvania developed ENIAC, the first general-purpose electronic digital computer. Using vacuum tubes to process the data, ENIAC executed 5,000 additions per second. Vacuum tubes are used to control electrons.

The advent of the transistor in 1947 changed everything. Starting in the 1950s, transistors replaced vacuum tubes in many systems, and enabled faster computers.

Meanwhile, in 1964, now-defunct Control Data introduced the CDC 6600, the world’s first supercomputer. Based on transistors, the 6600 incorporated a 60-bit processor with 2 MIPS of performance.

Fast forward to today, and the smart phone is faster than the early computers. Apple’s iPhone 12 incorporates the A14 processor based on TSMC’s 5nm process. Incorporating 11.8 billion transistors, the A14 features a 6-core CPU and a 16-core neural engine capable of 11 trillion operations per second.

At the high end, Fugaku in 2021 retained its position as the world’s fastest supercomputer. Built by Riken and Fujitsu, Fugaku is based on Arm’s A64FX processor. It has 7,630,848 cores, enabling 442 petaflops per second of performance. A petaflop performs one quadrillion floating-point operations per second.

Fugaku is in operation and is being used for various research projects. “(Fugaku) embodies technologies realized for the first time in a major server general-purpose CPU, such as 7nm process technology, on-package integrated HBM2, terabyte-class streaming capabilities and an on-die embedded high-performance network,” said Satoshi Matsuoka, director of the Center for Computational Science at Riken, in a paper at the 2021 Symposia on VLSI Technology and Circuits.

“We are well into the petaflop computing era,” said Aki Fujimura, CEO of D2S. “There are many research computers around the globe that are approaching exascale computing (1,000 petaflops). We will have many exascale computers by the end of this decade.”

Indeed, the industry requires more compute power to solve current and future problems in biotechnology, defense, materials science, medicine, physics, and weather prediction.

“We need to compute more at the same price. The problems are getting harder. The problems we serve are getting bigger and harder on top of that,” Fujimura said.

While traditional computing will continue to progress, the industry is rushing to develop quantum computing. In theory, these new systems promise to outperform today’s supercomputers, which could speed up the development of new technologies.

In the distant future, quantum computers are expected to be able to crack the world’s most complex algorithms within a reasonable time. This includes Shor’s algorithm, an integer factorization problem that can be utilized to break the widely used public-key cryptography scheme known as RSA.

Conceived in the 1980s, quantum computing has made some major strides over the years. Recently, two systems have achieved “quantum supremacy.” This describes a point where quantum computers can do things that a classical computer can’t.

Still, quantum computing is in its infancy. Work is underway to advance these systems, and find useful applications for the technology. “All systems that exist today are primarily used to explore future quantum applications, including looking at variational quantum algorithms for quantum chemistry, and quantum kernel estimation methods for machine learning,” said Jerry Chow, director of quantum hardware system development at IBM. “The systems that are deployed today are also interesting from the standpoint of benchmarks and characterization of their own performance, and to understand underlying noise sources to improve future iterations of these systems. One other aspect is to explore the concept of quantum error correction.”

Even if quantum computers realize their potential, they won’t replace today’s computers. “Quantum computing is clearly an important future technology for some types of computing problems. Prime factorization is another task that quantum computing is known to be far superior at than classical computing,” D2S’ Fujimura said. “In a way, quantum computing will augment classical computing for some specific difficult problems. On a larger scale, quantum computing will not replace classical computing. Classical computing is more appropriate for many of the tasks we need to compute.”

Today’s quantum computers are different and resemble giant chandeliers. These systems are housed in a dilution refrigerator capable of shielding the processor and other components from external noise and heat. The unit cools the devices between 10 and 15 milliKelvin.

Fig. 1: IBM’s Quantum System One line of quantum computers Source: IBM

Fig. 2: Researchers adjusting the dilution refrigerator in Intel’s lab. Source: Intel

A quantum system consists of a processor, which incorporates the qubits. Those qubits come in two configurations, with one-qubit and two-qubit gates. Let’s say you have a quantum processor with 16 qubits. The qubits are arranged in a two-dimensional 4 X 4 array. The first three rows (top to bottom) may consist of one-qubit gates. The last row may have two-qubit gates.

The processing functions are complex. In classical computing, you put a number into the computer, it calculates the function, and gives you an output.

Let’s say you have problem with 2bits of data. “If you have ‘n’ bits, you have 2n. That’s an exponentially large number of states, and you can only work on them one at a time. So, it’s exponential time or exponential in space,” explained William Oliver, a professor at the Massachusetts Institute of Technology (MIT), in a video presentation. “A quantum computer, on the other hand, can take those 2different components and put them all into one superposition state simultaneously. And this is what underlies the exponential speed up that we see in a quantum computer.”

There are other advantages. “In order to double the power of a quantum computer, you only have to add one qubit. It’s exponential. In order for a quantum computer to keep up with a classical computer in terms of Moore’s Law, they only have to add one qubit every 24 months,” said Paul Smith-Goodson, an analyst at Moor Insights & Strategy.

This all works in theory. What prevents quantum computing from realizing its full potential are several major issues. First, qubits lose their properties, typically within 100 microseconds, due to noise, according to IBM.

That’s why qubits must operate in extremely cold environments. “Qubits are extremely sensitive to their environment,” FormFactor’s Leong said. “Quieting down the qubit environment in a very cold or cryogenic environment is critical.”

In addition, noise causes errors in the qubits. So quantum computers require error correction. On top of that, the industry needs to scale up quantum computers with thousands of qubits. It’s nowhere close to that figure.

All told, quantum computing requires some breakthroughs. “We need to make qubits better than we’re making them today. And that’s across the field,” Intel’s Clarke said. “To me, the biggest challenge is how you wire them up. Every qubit requires its own wire and its control box. That works well when you have 50 or 60 qubits. It doesn’t work well when you have a million of them.”

Manufacturing qubits with good yields is also critical. Onto Innovation and others are developing metrology processes around the technology.

“Right now, we’ve conducted measurements on a few wafers or coupons,” said Kevin Heidrich, senior vice president at Onto. “The key behind most of the foundational technologies in quantum is utilizing the manufacturing technologies developed for classical computing. However, many are tweaking the devices, designs and integrations to enable quantum/qubit devices. The key engagements we have are around enabling precise and characterized devices to enable various forms of quantum computing such as photonic or spin qubits. Our focus is to provide metrology solutions to enable our development partners to best characterize their early devices, including things like precise sidewall control, materials thickness, and interface quality.”

Superconducting qubits
Today, there are 98 organizations working on quantum computers and/or qubits, according to the Quantum Computing Report. Companies are developing different types of qubits, including ion trap, neutral atoms, photonics, silicon spin, superconducting and topological. Each type is different, with some advantages and disadvantages. It’s too early to say which technology is superior.

“We really don’t know which technology is going to be the right technology to build a grand scheme fault tolerant machine. Companies have a five-year roadmap, leading to where they are going to have enough qubits to actually do something meaningful,” said Smith-Goodson from Moor Insights & Strategy. “(Regarding the installed base), IBM has a large number of machines. They have over 20 quantum computers and no one can match that. They have a large ecosystem built up around it. They have a lot of universities and companies that they’re working with.”

So far, superconducting qubits have made the most progress. In this category, D-Wave has gained attention by using quantum annealing, a technology that solves optimization problems. For example, if you have a problem with many combinations, a quantum annealing system searches for the best of many possible combinations. These capabilities have been demonstrated, at least to some degree.

Most of the activity is taking place in the bonafide quantum computer market using supercomputing qubits. Google, IBM, Intel, MIT, Rigetti, USTC and many others are developing products here.

Superconducting qubits are built around Josephson junctions. A Josephson junction includes a thin insulating layer, which is sandwiched by two superconducting metals. In operation, electrons pair up and tunnel through the junction.

In 2014, IBM demonstrated a 3-qubit device. Today, IBM sells a quantum computer with 65 qubits. Until recently, IBM led the industry in terms of overall qubit count in the superconducting space, according to the Quantum Computing Report. At present, the unofficial record is held by USTC with 66 qubits. IBM is next with 65, followed by Google with 53 qubits, Intel (49) and Rigetti (32), according to the Quantum Computing Report.

Qubit count isn’t the only factor. They also must have relatively long coherence times and gate fidelities. “Qubits and quantum processors are the central part of quantum hardware,” IBM’s Chow said. “To build a quantum computer or a quantum computing system, we will need not only quantum hardware, but also control electronics, classical computing units, and software that runs quantum computing programs.”

On that front, IBM offers Qiskit, an open-source quantum software development kit. “Our goal is to have a broad engagement of the developer community and grow a quantum ecosystem to bring quantum computers to users as their essential tools in their research and business,” Chow said.

The industry also will require systems with thousands of qubits, but vendors have a long way to go here. The results are still promising, however. In 2019, Google’s 53-qubit processor, called Sycamore, completed a calculation in 200 seconds. Google claimed it would take a supercomputer about 10,000 years to finish the same task.

Then, in June of 2021, China’s USTC presented a paper on Zuchongzhi, a 66-qubit superconducting quantum processor. In a calculation, USTC utilized 56 qubits. It performed a task 2 to 3 times faster than Google’s 53-qubit processor. “We expect this large-scale, high-performance quantum processor could enable us to pursue valuable quantum applications beyond classical computers in the near future,” said Jian-Wei Pan, a professor of USTC, in a paper. Others contributed to the work.

The results from China and elsewhere are up for debate. Many don’t use any benchmarks to report their results, including quantum volume, which is a metric to express the effectiveness of a quantum computer. “It all doesn’t depend on qubits. We don’t know how many of these systems perform. If you don’t have error correction and get up to a certain point, you can add all the qubit you want to and it’s never going to be any more powerful,” said Smith-Goodson from Moor Insights & Strategy.

Meanwhile, besides USTC’s processor, there are other developments in superconducting qubits:

  • Rigetti introduced a multi-chip quantum processor, enabling an 80-qubit system by year’s end.
  • By year’s end, IBM will release Eagle, a 127-qubit quantum processor. IBM is working on a 433-qubit processor for 2022, and a 1,121-qubit device for 2023.
  • Google found a way to reduce qubit error rates. It also plans to develop a 1 million qubit processor by 2029.

Ion traps
Ion trap qubits are another promising technology. With ion trap, atoms are at the heart of the quantum processor. The atoms are trapped, and then lasers do everything from the initial preparation to final readout, according to IonQ, a developer of the technology.

In ion trap, IonQ is leading with 32 qubits, followed by AQT (24), Honeywell (10) and others, according to the Quantum Computing Report.

On the R&D front, Sandia National Laboratories is developing QSCOUT, a quantum computer testbed based on ion trap qubits. QSCOUT is a 3-qubit system. Sandia plans to expand the system to 32 qubits over time.

With QSCOUT, Sandia is offering an open-access program for end-users. “Not only can users specify which gates (each circuit is made up of many gates) they want to apply and when, but they can also specify how the gate itself is implemented, as there are many ways to achieve the same result. These tools allow users to get into the weeds of the how the quantum computer works in practice to help us figure out the best way to build a better one,” said Susan Clark, a physicist and the QSCOUT lead at Sandia.

“Since we are a testbed system, the code running on our machine is generated by users, who have lots of ideas of what they might like to run on a quantum computer,” Clark said. “Thirty-two qubits are still small enough that it can be fully simulated on a classical computer, so the point is not to do something that a classical computer cannot do. The main reasons for building the smaller system are: 1) study how to map problems onto a quantum computer the best way for best performance on a future larger system (quantum chemistry, quantum system simulations), and 2) learn techniques for making a quantum computer run better that can be applied to a bigger machine.”

Like the superconducting qubit market, ion trap is also seeing a wave of activity. Honeywell, for example, is spinning off its quantum computing unit and will merge it with Cambridge Quantum Computing. Honeywell also demonstrated the ability to correct quantum errors in real time.

IonQ’s customers, meanwhile, can purchase access to its quantum computers via Google’s cloud services.

Silicon qubits
Silicon spin qubits are also promising. Leti, Intel, Imec and others are working on this technology. Intel appears to be leading with 26 qubits, according to the Quantum Computing Report.

“What we’re doing here is making a single electron transistor,” Intel’s Clarke said. “We’re making a transistor that has one electron in the channel. That single electron can either have spin up or spin down. That spin up or spin down represents the ‘0’ and the ‘1’.”

The trick is to make the electron move into the superposition state. “When you have one spin, it’s one qubit,” Clarke said. “If you have two electrons close to each other, or two of these spin qubits, then you can start to perform operations. You can start using quantum entanglement.”

Silicon spin qubits have some advantages. “Intel’s spin qubits are a million times smaller than some of the other qubit technologies,” Clarke said. “We’re going to need 100,000 to 1 million of them. When I envision what a quantum chip will look like in the future, it will look similar to one of our processors.”

In addition, spin qubits leverage some of the same processes and tools used in semiconductor fabs. The processes don’t involve leading-edge nodes. “A lot of our innovation comes more from the materials that we’re using rather than the patterning capability,” Clarke said.

Fig. 3: Spin qubit chip placed on the tip of a pencil eraser Source: Walden Kirsch/Intel

There is a frenetic among of activity in silicon spin:

  • Intel rolled out Horse Ridge II, a second-generation cryogenic control chip. The device brings control functions for quantum computer operations into the cryogenic refrigerator, which can streamline the complexity of the control wiring for quantum systems.
  • CEA-Leti has developed an interposer that enables the integration of devices for quantum computing. The interposer connects qubits and control chips.
  • Imec devised uniform spin qubit devices with tunable coupling in a 300mm integrated process.
  • Intel and FormFactor have separately developed cryoprobers. These systems characterize qubits at cryogenic temperatures.

There are other types of qubits, as well. “You have photonics. People are using light particles and that looks like a promising field,” said Smith-Goodson from Moor Insights & Strategy.

But it’s unclear which technologies will prevail over time. The same is true for companies in this space.

Perhaps a bigger question is whether quantum computing will ever live up to the hype. But companies and countries are betting big on this technology. And given the progress so far, the current results and activity make this all worth watching.

The Race To Make Better Qubits
How dopant atoms could make qubits that last much longer.
The Chip Industry’s Next-Gen Roadmap
SRC’s new CEO sheds some light on next-gen projects involving everything from chiplets to hyperdimensional computing and mixed reality.
The Long Road To Quantum Computing
Focus shifting from novel techniques and materials to what the chip industry knows best – silicon.
Quantum Issues And Progress (2019)
Work begins on building a quantum computing ecosystem.
Security Implications Of Quantum Computing
The race is on to find and implement a public-key cryptographic algorithm that will stand up to the challenges posed by quantum computers.

Leave a Reply

(Note: This name will be displayed publicly)