IBM sets 2026 target for quantum advantage, fault-tolerant quantum computing by 2029

IBM announced during the 2025 Quantum Developer Conference on Wednesday that it expects to reach quantum advantage by 2026, while targeting full-scale fault-tolerant quantum computing by 2029.

The company confirmed that its Nighthawk processor, a chip with major architectural upgrades, will start reaching users this year.

Jay Gambetta, who leads IBM Research as Director and also holds the title of IBM Fellow, said, “We believe that IBM is the only company that is positioned to rapidly invent and scale quantum software, hardware, fabrication, and error correction to unlock transformative applications.” Jay added that the company is announcing several critical milestones on its roadmap today.

Nighthawk, now revealed as IBM’s most advanced chip yet, is built with 120 qubits and uses 218 tunable couplers to link each qubit to its four nearest neighbors. That’s a 20% increase in coupler count compared to the Heron processor, according to an official press release issued by IBM.

The new setup allows researchers to run 30% more complex circuits while keeping the error rates low, a crucial requirement to perform up to 5,000 two-qubit gates in a single job.

IBM targets 15,000 two-qubit gates by 2028

Deliveries of Nighthawk will begin before the end of 2025. But this chip is just the start. IBM plans to push performance even further. By the end of 2026, it expects to boost capacity to 7,500 gates, jumping to 10,000 in 2027, and 15,000 in 2028.

Those future versions will integrate more than 1,000 connected qubits using long-range couplers, a system tested last year on internal experimental processors.

As it builds this pipeline, IBM is pushing for community-driven validation. It launched a quantum advantage tracker, developed with help from Algorithmiq, the Flatiron Institute, and BlueQubit, to measure and verify progress in real time.

The tracker already includes three experiments testing quantum advantage in observable estimation, variational algorithms, and classical-verifiable problems.

Sabrina Maniscalco, the CEO and co-founder of Algorithmiq, said, “The model we designed explores regimes so complex that it challenges all state-of-the-art classical methods tested so far.”

She said early results look promising, with Flatiron Institute confirming that the results are hard to simulate on classical systems.

Hayk Tepanyan, the co-founder and CTO of BlueQubit, added that his team is focused on tracking quantum workloads where classical machines are already starting to fall behind.

“Through our work around peaked circuits, we are excited to help formalize instances where quantum computers are starting to outperform classical computers by orders of magnitude,” he said.

Qiskit boosts error control with C-API and dynamic circuits

To support this push, IBM is scaling up its software. The Qiskit stack now supports dynamic circuit capabilities that increase output accuracy by 24% on jobs involving 100 qubits.

It also now supports a new execution model using a C-API, letting developers integrate with HPC environments and use them to reduce error correction costs by more than 100 times.

IBM is also releasing a C interface for Qiskit so users can run quantum workloads inside existing high-performance compute systems.

The company said that by 2027, the Qiskit stack will add computational libraries focused on machine learning and optimization. These tools will help address physics and chemistry problems like differential equations and Hamiltonian simulations.

The company also revealed that it is actively building toward fault-tolerant quantum computing on a parallel track. Its new Loon processor, which was also announced during the event, includes all the components needed to demonstrate efficient, scalable quantum error correction.

It includes multi-layer routing that links qubits across longer distances with “c-couplers,” and enables qubit resets between operations on the same chip.

To top that off, IBM confirmed it can now decode quantum errors in under 480 nanoseconds using qLDPC codes, running entirely on classical hardware. That engineering achievement came a full year ahead of schedule.

Together with Loon, it lays the foundation for scaling qLDPC across fast, high-fidelity superconducting qubit systems, the same qubits used in all of IBM’s hardware.

Production of IBM’s quantum processor wafers has now shifted to a 300mm fabrication facility at the Albany NanoTech Complex in New York. This shift allows for faster chip iteration and more complexity.

IBM said it has already cut development time in half and increased the complexity of its quantum chips tenfold using the new equipment.

It’s also now able to explore multiple processor designs simultaneously, helping push both the Nighthawk and Loon platforms forward at once, according to the press release.

If you're reading this, you’re already ahead. Stay there with our newsletter.