IBM has announced what it claims to be a significant breakthrough in quantum computing architecture today, solving the scientific obstacles to error correction. The company said that it is now on the path to building the world’s first scalable, error-corrected quantum computer by 2029.
“We’ve cracked the code to quantum error correction,” said Jay Gambetta, vice president at IBM Quantum, at a press conference on Monday.
The new quantum computer, named IBM Quantum Starling, will be built at the IBM quantum computing data center in Poughkeepsie, N.Y., and will have 200 logical qubits, which translates to about 20,000 physical qubits.
“Starling will be capable of running 100 million quantum operations,” Gambetta told reporters. To match the computational state of the Starling would require the combined memory of a quindecillion of today’s most powerful supercomputers, he adds. “That’s a one followed by 48 zeros.”
Starling will be followed in 2033 by Blue Jay, which will be ten times larger.
The breakthrough was made possible by the fact that IBM abandoned its previous surface code approach to error correction for low-density parity check (LDPC) codes. IBM published its research in a paper in March of 2024. The new approach was ten times more efficient than previous methods, the company said.
Error correction is the biggest obstacle to practical quantum computing. Quantum computing companies typically address error correction with redundant qubits, but with previous approaches the number of these redundant qubits would grow much faster than the number of total usable qubits, keeping the computers from reading any useful size.
Earlier this month, IBM published two more papers, demonstrating the value of its new approach to quantum correction.
“In the first paper, we show how the qubit advantage is retained even when we build large logical circuits on modular quantum system architecture,” says Matthias Steffen, IBM’s head of quantum process technology. “In the second paper, we show how we identify and correct errors in real time using conventional computing resources. When taken together, these papers will demonstrate the essential criteria for a large-scale error correction approach.”
“With this news, we’re confident that large-scale quantum computing is no longer a question of science but an engineering challenge,” says Gambetta.
However, even Blue Jay’s 2,000 logical qubits may not be enough to make quantum computing a practical reality for businesses. According to quantum computing expert Bob Sutor, founder and CEO at Sutor Group Intelligence and Advisory, it will take 100,000 qubits before quantum computers get interesting.
A shorter-term goal is quantum advantage, where a quantum computer can perform a calculation impossible or impractical for today’s classical computer.
According to Gambetta, IBM will be able to demonstrate quantum advantage “in the near future.” Another company, D-Wave Quantum, has already claimed to have achieved that goal in March.
D-wave’s approach uses a very different quantum architecture, called annealing, which is more akin to an analog computer. Annealing quantum computers aren’t a general-purpose computer like those that IBM and other quantum computing companies are building.
And, according to Gambetta, some of IBM’s collaboration partners using its quantum computing platforms are already getting results beyond what is possible with annealers.
The company also announced some intermediate steps on the road to Starling. Later on this year, we will see the release of the IBM Quantum Loom computer, which will test the architectural components of the new quantum error corrections. Then, next year, IBM will release Kookaburra, which will combine quantum memory with logic operations, a basic building block for scaling quantum computers beyond a single chip.
Today’s announcement is “a significant leap in a long marathon ahead,” says Gartner analyst Chirag Dekate. Enterprises may see accelerated development in areas requiring complex computations, he says, such as quantum chemistry and advanced simulations. “The improved error correction and lower qubit overhead could make quantum computing more accessible for practical applications,” he adds.
Dekate points out that Google has also announced progress in quantum correction with its Willow chip. Google also plans to have a useful quantum computer by 2029, he adds.
Meanwhile, it’s still not clear as to how quantum computing can translate into tangible business value, Dekate says. “The reality in quantum is that we are not yet at the ChatGPT-like moment,” he says.