IBM this week reiterated that it plans to have a “fault-tolerant” quantum computer ready by 2029. Though the statement was in line with IBM’s previously-stated quantum computing roadmap expectation for achieving error correction in a quantum computer, company officials said they have “de-risked” their path to this achievement, giving them more certainty they will reach it on time.
The planned system, called IBM Quantum Starling, will be built in a new IBM Quantum Data Center in Poughkeepsie, New York. It is expected to be capable of running 100 million quantum operations using 200 logical qubits, equating to about 20,000 times more operations than today’s quantum computers, according to. The error correction methods IBM is using could require up to 2,000 physical qubits to achieve that many logical qubits, so the company still has a long way to go.
But Jay Gambetta, IBM Fellow and vice president of IBM Quantum, said in a briefing that his team’s recent engineering studies have lit the path to show how IBM will get there, and that the company is now “more and more certain” of how and when it will have a fault-tolerant quantum computer. “I would not call this an acceleration of our roadmap, but we have de-risked the problems we faced to get there,” he said.
Matthias Steffen, IBM Fellow, Quantum Processor Technology, added that IBM’s studies, including two papers just released, help demonstrate “the essential criteria for a large-scale error correction approach,” including:
- Fault tolerant: Logical errors are suppressed enough for meaningful algorithms to succeed.
- Addressable: Individual logical qubits can be prepared and measured throughout the computation.
- Universal: A universal set or arbitrary unitary operations can be applied to the logical qubits.
- Adaptive: Measurements are real-time decoded and also subsequent quantum instructions.
- Modular: The hardware can be distributed across a set of connected, replaceable modules.
- Efficient: Meaningful algorithms can be executed with reasonable resources.
By those measures, IBM thinks and claims it will have the first large quantum computing system with error correction, though many other companies are pursuing that goal at the same time, each back by their own theories, studies, and experiments, including Amazon, Google, and Microsoft.
One of the papers mentioned above outlines how quantum low-density parity check (qLDPC) codes can be used to process instructions and run operations on an error-corrected system. “This code drastically reduces the number of physical qubits needed for error correction and cuts required overhead by approximately 90%, compared to other leading codes,” according to a company statement. “Additionally, it lays out the resources required to reliably run large-scale quantum programs to prove the efficiency of such an architecture over others. The second paper describes how to efficiently decode the information from the physical qubits and charts a path to identify and correct errors in real-time with conventional computing resources.”
Starling is not the only milestone on IBM’s roadmap. Leading up to 2029, the company also is planning IBM Quantum Loon, which is expected later this year, and is designed to test architecture components for the qLDPC code, including “C-couplers” that connect qubits over longer distances within the same chip. Next year will feature IBM Quantum Kookaburra, which will be IBM’s first modular processor designed to store and process encoded information. It will combine quantum memory with logic operations, the basic building block for scaling fault-tolerant systems beyond a single chip. IBM Quantum Cockatoo, expected in 2027, will entangle two Kookaburra modules using “L-couplers.” This architecture will link quantum chips together like nodes in a larger system, avoiding the need to build impractically large chips, the company said.
Gambetta and Steffen said the Starling system will be the foundation for IBM Quantum Blue Jay, the next machine on the roadmap, which will be capable of executing 1 billion quantum operations over 2,000 logical qubits.