IBM has updated its quantum computing roadmap, with plans to roll out a new system called Starling by the end of the decade. If all goes as planned, Starling could handle far more operations than current machines and push quantum computing closer to real-world use.
Starling is designed to carry out 100 million quantum operations using 200 logical qubits. It is to be hosted in a new IBM data centre under construction in Poughkeepsie, New York. Renders of the facility suggest it will hold four of IBM’s Quantum System Two machines – each using a hexagonal layout that fits three quantum processors – and sit alongside rows of standard server racks.
Another machine, Blue Jay, is intended to follow Starling. It is said to be able to support 2,000 logical qubits and perform 1 billion operations, according to IBM. Both Starling and Blue Jay would mark a shift from earlier designs which are based on single, monolithic processors.
IBM has spent years trying to scale up quantum systems. But researchers have found that simply packing more qubits on one chip wasn’t enough to build fault-tolerant machines. To reach better levels of reliability, IBM is now focusing on modularity – a design approach that links smaller quantum units into one system. The setup could help avoid some of the limits of older surface code methods.
The company also says fault tolerance isn’t just about qubits. The whole system – including gates, connections, operations, control electronics, memory, and how errors are measured – has to be reworked from the ground up.
IBM says Starling’s performance would be beyond what classical supercomputers can achieve. The company says the computational state of Starling would require more memory than all of the world’s most powerful supercomputers combined – although no technical breakdown was shared to support this.
CEO Arvind Krishna says the company’s focus is now on building machines that can take on real-world problems. That means reaching a point where quantum computers can run tasks that classical computers simply can’t handle, even with massive amounts of time or power.
The isn’t IBM’s first quantum computing roadmap. Back in 2020, it laid out a plan that included the Falcon processor (27 qubits), and later the 1,121-qubit Condor chip, which was aimed at research use but was never released to the public. Since then, it has become clear that bigger chips alone won’t solve the problem of making quantum systems stable and usable.
IBM’s new approach includes a series of processors with different features. Nighthawk, expected this year, will offer 133 error-corrected qubits. Multiple Nighthawk units can be connected to reach over 1,000 qubits total. IBM plans to launch Loon in 2025, Kookaburra in 2026, and Cockatoo in 2027. Each step adds new features: longer qubit connections, combined logic and memory, and a way to link chips together like nodes in a larger system.
Error correction remains one of the toughest problems in quantum computing. IBM says it’s made some headway with quantum low-density parity check (qLDPC) codes. The technique may cut down how many physical qubits are needed to support each logical qubit, and IBM claims it could lower the overhead by up to 90 per cent.
The company is also working on ways to catch and fix errors in real time, saying this can be done with standard computing hardware like FPGAs or ASICs. This too would mark a shift from current methods, which often require complex workarounds after errors occur.
Jay Gambetta, a longtime IBM researcher and VP of IBM Quantum, told reporters the company has “cracked the code” for error correction, but the rest of the engineering still needs work.
IBM started its quantum research in the late 1990s, teaming with universities to build early test systems. In 2016, it launched a five-qubit machine online through IBM Quantum Experience. Since then, the field has seen steady improvements, but no system yet has crossed into fully fault-tolerant territory.
Whether IBM can meet its 2029 target remains to be seen. But the company’s new roadmap shows a shift in focus – from adding more qubits to a single to making quantum systems stable, connected, and useful via modularity.