IBM Starling: 20,000X Faster Than Today's Quantum Computers

Forbes

10-06-2025

IBM quantum computers
Today IBM released a roadmap to Starling, a quantum computer with 20,000 times the processing power of today's quantum computers. Starling won't be built until 2029, but IBM says they've cracked the toughest problems on the path, and that this roadmap is trustworthy. A key breakthrough: 14X better error correction, which solves one of the most challenging problems in quantum computing: quantum decoherence.
IBM will have a fully fault tolerant large-scale quantum computer by 2029, IBM fellow and director of quantum systems Jerry Chow told me on the TechFirst podcast.
'We really have a path to make this viable in this timescale."
IBM is aiming high. Until today, the company says, a clear path to building a large-scale fault-tolerant quantum computer without unrealistic engineering overhead has not been published. Starling will be such a computer, and Blue Jay, the next quantum computer in IBM's roadmap, will have 2,000 logical qubits, and could run a billion quantum operations effectively instantly.
'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,' says Arvind Krishna, Chairman and CEO of IBM.
One of the key innovations is better error correction. Today's quantum computers can require 10,000 physical qubits to form a single logical qubit that is fault-tolerant enough to run meaningful operations. IBM's new error correction system, first unveiled in Nature magazine in 2024, reduces the number of physical qubits required by about 90%. The result is that instead of needing millions of physical qubits for a useful quantum computer, IBM is aiming to achieve a ratio of hundreds or perhaps thousands of physical qubits per logical qubit.
That's a massive improvement: even a quantum leap, if the company can pull it off.
IBM's path to Starling is iterative and grounded in hardware milestones:
Another innovation is a novel way of interlinking qubits in a three-dimensional matrix that Chow likened to a physical neural network, where qubits are connected like neurons in a brain, enabling richer and more scalable interactions.
IBM's new quantum computers will use a much more complex 3D lattice connection that increases the ... More number of connections between qubits.
In addition, Chow told me IBM is getting less and less bespoke in its processes for building quantum computers. That means the company is taking steps towards mass manufacturing, another key step to making quantum computers less science experiment and more standard engineering and production problem.
'I've been working with superconducting qubits since 2005, and it's always been a rather bespoke process of design,' he says. 'You simulate, you design, and you lay out by hand, and simulate … but then over time we've really developed a lot of the toolboxes that you'd need for advanced manufacturing methods, advanced simulation methods to really get parameters close to first-time-right.'
As projected, Starling will be so massively capable that to represent its computational state would require the memory of more than a quindecillion of the world's most powerful supercomputers, IBM says.
A quindecillion is a 1 followed by 48 zeros. To make that make sense, think of counting to one quindecillion. At one number per second, it would take longer than the age of the universe. It gets worse: if you had a quindecillion dollars, and spent a trillion dollars every second, you'd still be spending for longer than the universe has existed.
In other words, this is quantum supremacy: the point where a quantum computer can perform a calculation that is practically impossible for any classical computer to do in a reasonable amount of time.
That means Starling would enable transformative progress in domains like:
• Drug discovery
• Advanced materials
• Battery chemistry
• Optimization problems
Of course, to make that reality, there's at least four long years of hard work. Time will tell if IBM Quantum can deliver on this promised roadmap.