Sep 15, 2020 - Science

IBM releases roadmap to quantum computing "inflection point"

IBM cryosat refrigerator for cooling quantum computers

IBM's new cryostat refrigerator for cooling quantum computers. Photo: Chris Nay/IBM

IBM thinks it can deliver a quantum computer that would marry enormous computing power with a tough-to-achieve low rate of errors by 2023, per a technical timeline the company published today.

Why it matters: Companies, most notably IBM and Google, are investing heavily in a race to commercialize quantum computers that ultimately may be able to solve some problems much faster than a classical computer. IBM's stake in the ground represents what the firm says is an "inflection point" after which the rate of progress in the field will rapidly increase.

  • "If you are a business or institution that relies on computation as means of competitive advantage or security, you need to know it is coming and act on it," says Dario Gil, director of IBM Research.
  • He says the roadmap lays out the parallel progress required in software, cryogenics and other technologies to support quantum computing, and is "part of the equation of how the industry evolves" in terms of workforce development, investments, and infrastructure.

How it works: Unlike the bits in classical computers that can represent two states of information — represented as "0" or "1" in binary code — quantum bits, or qubits, can hold multiple states of information at once, a feature known as superposition.

  • That gives quantum computers the ability to simultaneously compute many potential solutions to a problem before delivering just one.
  • Qubits can be entangled — if the state of one qubit is changed, the state of its pair will also change, no matter how far apart they are.
  • Entanglement and superposition are exploited by quantum computers and give them their processing power, which increases exponentially with the number of qubits in the machine.

Yes, but: As the number of qubits increases in a system, so does the rate at which the computer makes errors.

  • "The biggest challenge facing our team today is figuring out how to control large systems of these qubits for long enough, and with few enough errors, to run the complex quantum circuits required by future quantum applications," IBM's Jay Gambetta wrote in a post announcing the roadmap.
  • Those errors could be caught and corrected by distributing the information a qubit encodes across many qubits.
  • IBM believes a quantum computer with 1,000 qubits or more will be able to perform a calculation a classical computer cannot but with a low enough error rate that it is reliably accurate.

Where it stands: Today's quantum computers have error rates much higher than classical computers because it is difficult to maintain a qubit's superposition state.

  • IBM's current 65-qubit quantum processor has an error rate of roughly 1.5% for entangling two qubits, pairs that serve as gates in the system, Gambetta, who leads IBM's quantum computing research, tells Axios.

What's next: IBM aims to build a 1,121-qubit system by the end of 2023. In the meantime, it believes it can achieve a 127-qubit machine by the end of 2021 and hit 433 qubits a year later.

  • "Innovations need to occur at all of these levels," says Gil, citing required advances in configuring and packaging qubits, controlling them, and hardest of all, reducing the error rate.

The intermediate goal is to get a 1,000-qubit machine to an error rate of 0.01%, says Gil. That would make roughly one error in every 10,000 operations (for classical computers, it is one in every 10^17). If that happens, "all of a sudden the quality of what is going to be possible to compute in a machine will no longer be about the hardware but the ability to innovate on the software," says Gil.

  • "A demonstration of that machine will tell us for certain that other world-changing machines will be possible," he says, referring to future quantum computers with potentially thousands or millions of qubits.

What they're saying: "If [researchers] can reach 1,000 qubits with an order of magnitude smaller error rates than today, then I think it could be a real inflection point," Arne Grimsmo, who studies quantum error correction at the University of Sydney, tells me in an email. 

  • "There are a whole host of challenges that need to be solved to get there, but if we can build a chip with 1,000 really high quality qubits, then it becomes believable that we could build a much larger chip with millions of qubits," he says.

Of note: IBM also announced it designed a "superfridge" for cooling large machines to near absolute zero — a key condition of quantum computing.

The big picture: The U.S. government and industry are ramping up investments in quantum computing and related sciences, including establishing five new quantum research centers announced last month.

  • The bigger picture: China and Europe are also pouring resources into quantum computing, an emerging technology with potential applications for cryptography, material science and other fields that is fueling the global tech race.

What to watch: Whether Google responds with its own timeline for benchmarking progress on quantum computers.

  • Last fall, Google announced it had achieved "quantum supremacy" — a demonstration that a quantum computer can perform a calculation that classical machines cannot practically compute.
  • IBM disputed the claim, both technically and philosophically.

Go deeper:

  • Why quantum computing matters (Axios)
  • The race to build a quantum economy (Axios)
  • The biggest flipping challenge in quantum computing (Science)

Editor's note: This story was updated with comment from University of Sydney researcher Arne Grimsmo.

Go deeper