IBM

Overview

IBM is one of the oldest technology companies in the world, and it has been building quantum computers for longer than almost anyone else in the industry. Its IBM Quantum division designs superconducting processors, builds the control electronics and cryogenic systems around them, and runs a cloud service that lets researchers, students, and companies run circuits on real hardware from anywhere.

What sets IBM apart is its habit of publishing a dated, public roadmap and then reporting against it. The current plan runs through a fault-tolerant machine called Starling, targeted for 2029, and extends to even larger systems in the 2030s. Because the milestones are public, it is unusually easy to check whether IBM is on schedule.

History and background

IBM's quantum story reaches back to the 1980s and 1990s, when IBM researchers such as Charles Bennett and Rolf Landauer helped lay the theoretical foundations of quantum information. The company moved from theory to public hardware in May 2016, when it put a small five-qubit processor on the cloud as the IBM Quantum Experience. It was one of the first times anyone outside a research lab could run a program on a real quantum chip. Qiskit, the open-source software development kit, followed in 2017.

Hardware then scaled quickly. IBM introduced its first integrated system for commercial use, IBM Q System One, in 2019. The Eagle processor passed 100 qubits in 2021, Osprey reached 433 qubits in 2022, and in December 2023 IBM unveiled the 1,121-qubit Condor alongside the 133-qubit Heron and a modular system called Quantum System Two. After Condor, IBM deliberately shifted its emphasis from raw qubit count to quality, modularity, and error correction.

How the technology works

IBM builds superconducting transmon qubits: tiny circuits made from superconducting metal that behave like artificial atoms when cooled to around 15 millikelvin inside a dilution refrigerator. Microwave pulses control each qubit, and qubits are coupled to their neighbors so they can run two-qubit gates. Superconducting qubits are fast, and they can be manufactured with techniques borrowed from the semiconductor industry, which is one reason both IBM and Google favor them.

Earlier IBM chips used a heavy-hex layout, where each qubit connects to two or three neighbors. That design limits certain kinds of noise but forces extra swap operations. The Nighthawk processor, delivered in November 2025, moves to a square lattice with 120 qubits and 218 tunable couplers, which lets circuits use fewer swaps and therefore run more complex programs before errors pile up.

For fault tolerance, IBM is betting on quantum low-density parity-check (qLDPC) codes rather than the surface code. IBM's published analysis suggests these codes could need roughly ten times fewer physical qubits for the same level of protection, but they require long-range connections between qubits. The experimental Loon chip, also delivered in November 2025, tests the extra wiring layers that make those connections possible. The Kookaburra module planned for 2026 is meant to be the first to store information in qLDPC memory and process it with an attached logical processing unit.

Products and platforms

Processors in our Hardware Database

• IBM Condor — 1,121 qubits, released 2023
• IBM Heron (r1, r2, and r3) — 156 qubits, released 2023 (r1); 2024 (r2); 2025 (r3)
• IBM Eagle — 127 qubits, released 2021
• IBM Osprey — 433 qubits, released 2022

Key milestones

Strengths and challenges

Strengths

• The most detailed and verifiable public roadmap in the industry, with named processors and target years.
• A large ecosystem: cloud access, the Qiskit SDK, and a broad network of research and enterprise partners.
• In-house expertise across chip design, cryogenics, control electronics, and software.
• A qLDPC error-correction strategy that, on paper, needs far fewer physical qubits than the surface code.

Challenges and open questions

• Fault tolerance is still years away: Starling is targeted for 2029, and Kookaburra's 2026 delivery had not been announced as of late September 2026.
• qLDPC codes need long-range connections that are hard to fabricate, and they are less proven experimentally than the surface code.
• Superconducting qubits need dilution refrigerators, and scaling wiring and cooling to very large systems is an unsolved engineering problem.
• Claims of quantum advantage on today's noisy chips are often challenged by improved classical algorithms, so results need independent verification.

Why it matters

For anyone learning quantum computing, IBM is the easiest place to begin. Access to real hardware, a mature Python SDK, and extensive documentation mean that you can write a circuit in the morning and see it run on a physical chip the same afternoon. Much of the vocabulary used across this site, from heavy-hex lattices to logical qubits, comes from IBM's public materials.

For the industry, IBM's roadmap acts as a benchmark. When a company promises a fault-tolerant machine, IBM's schedule gives everyone else something concrete to compare against. Whether IBM meets its 2026 and 2027 milestones will be one of the clearest signals about how quickly error-corrected quantum computing is actually arriving.

Latest news

• Kookaburra, IBM's first qLDPC memory module, remains on the 2026 roadmap; no delivery had been announced as of late September 2026 — September 2026
• IBM delivers the 120-qubit Nighthawk and the experimental Loon processor — November 2025
• IBM says it expects the first demonstrations of quantum advantage by the end of 2026 — November 2025
• IBM publishes its roadmap to the Starling fault-tolerant system planned for 2029 — June 2025

Frequently asked questions

This profile is written for educational purposes and summarizes public announcements, company filings, and published research. Last reviewed: September 30, 2026. It is not investment advice. Check the company's official channels for the latest figures and claims.

← Back to all quantum companies