IBM has updated its quantum computing roadmap, claiming it will be able to offer hardware more powerful than any classical silicon systems before the end of the decade.

The company this week released a roadmap that it says will allow it to build the “world’s first large-scale, fault-tolerant quantum computer, setting the stage for practical and scalable quantum computing.”

In IBM parlance, utility-scale quantum computing means running workloads that cannot be simulated via brute force on classical techniques.

By 2029, IBM aims to deliver Starling, a new large-scale quantum computer that will be housed in a new IBM quantum data center in Poughkeepsie, New York, and is expected to perform 20,000 times more operations than today’s quantum computers.

Details on the data center are sparse, but in a video the company said building work has started and is ongoing. A rendering suggests the facility will feature four separate IBM Quantum System Twos, featuring its unique hexagonal design housing three quantum processor units (QPUs) each, with multiple conventional racks alongside.

Starling would appear to combine six connected hexagonal quantum systems alongside two rows of conventional racks. Blue Jay appears to have a similar design to Starling, but comprises 11 linked hexagonal systems in the render.

“To represent the computational state of an IBM Starling would require the memory of more than a quindecillion (ten to the power of 48) of the world’s most powerful supercomputers,” the company claimed.

“IBM is charting the next frontier in quantum computing,” said Arvind Krishna, chairman and CEO, IBM. “Our expertise across mathematics, physics, and engineering is paving the way for a large-scale, fault-tolerant quantum computer — one that will solve real-world challenges and unlock immense possibilities for business.”

Starling will reportedly be able to run 100 million quantum operations using 200 logical qubits. It will be a precursor to Blue Jay, another quantum system capable of executing 1 billion quantum operations over 2,000 logical qubits.

The updated roadmap includes a number of new processors. The first version of the upcoming Nighthawk QPU, coming this year, will offer 133 error-corrected qubits, with linked versions offering up to 1,080 qubits with a larger number of gates.

IBM then aims to release the Loon processor, expected in 2025; Kookaburra, expected in 2026; and Cockatoo, expected in 2027. The new chips will enable the company to connect qubits over longer distances within the same chip, then combine quantum memory with logic operations, and finally link quantum chips together like nodes in a larger system and avoid the need to build impractically large chips.

Big Blue is publishing two new research papers outlining some of the advancements it has made around creating more powerful quantum systems. The company details the quantum low-density parity check (qLDPC) codes that it says reduces the number of physical qubits needed for error correction and cuts required overhead by approximately 90 percent. It also outlines a path to identify and correct errors in real-time with conventional computing resources – FPGAs or ASICs.

In a press briefing, IBM fellow and VP of IBM Quantum, Jay Gambetta, said the company had “cracked the code” to quantum error correction, but there was “a lot of work” still to be done on the engineering side.