Industry & Milestones

IBM Heron & the 100,000-Qubit Roadmap: What IBM Quantum's Latest Chips Tell Us

Silicofeller Research TeamPublished by Silicofeller · Industry Analysis · Jun 2026

IBM's Heron processor, released in late 2023 and now the backbone of IBM Quantum's cloud fleet, introduces a fundamentally new coupler architecture that addresses the largest source of error in previous IBM processors: always-on ZZ crosstalk between neighbouring qubits.

"Eagle and Osprey used fixed-frequency transmons with cross-resonance gates. Every qubit was permanently coupled to its neighbours, creating a constant background of ZZ interactions that limited circuit depth. Heron replaces this with tunable couplers that switch the interaction on and off — reducing residual ZZ by more than 100×."

The Tunable Coupler Architecture

Heron uses a direct microwave interaction between fixed-frequency qubits mediated by a tunable coupler element. When the coupler is biased to its operating point, it enables fast two-qubit gates (~100 ns). When idle, the coupling is suppressed to below 1 kHz — compared to 10–100 kHz residual ZZ in previous generations.

ProcessorYearQubits2Q Gate ErrorResidual ZZ
Eagle r32023127~0.3%~50 kHz
Osprey2022433~0.5%~80 kHz
Heron r12023133~0.1%<500 Hz
Heron r22024133~0.05%<200 Hz

IBM's 2033 Roadmap

IBM has published an aggressive roadmap targeting a 100,000-qubit "Quantum Centric Supercomputer" by 2033. The architecture uses a modular design where multiple Heron-class chips are connected via quantum communication links, with classical HPC resources tightly integrated for error correction decoding. Key milestones include a 1,000-logical-qubit system by 2028 and demonstration of quantum advantage on a commercially relevant problem by 2026.

Key Takeaways

  • Heron's tunable coupler architecture reduces residual ZZ crosstalk by over 100× compared to Eagle, enabling much deeper circuits.
  • IBM is pursuing a modular, multi-chip path to scale — rather than building ever-larger monolithic processors.
  • The 100,000-qubit target by 2033 is ambitious but grounded in a credible architectural roadmap.
  • Heron's 2Q gate fidelity of 99.95% is now competitive with the best trapped-ion processors.

About the Authors

SF

Silicofeller Engineering Team

The Silicofeller team specialises in superconducting quantum chip design automation, electromagnetic simulation, and VLSI-grade layout tooling. Our mission is to make quantum hardware design accessible, reproducible, and physics-grounded.