Quantum roadmaps are easy to misread. Vendors headline qubit counts and future dates, but the real race runs on a different metric: logical qubits, the error-corrected units that hold up through a long computation. Reading the roadmaps correctly means watching logical qubits, not physical ones.
IBM: Starling by 2029
IBM's plan targets a machine called Starling in 2029, built from modular cryogenic cells linked together. In August 2026 IBM connected its first modular cryogenic cells, the step that lets fridges share qubits and scale beyond one box. The roadmap runs Heron today, then larger modular systems, then fault-tolerant Starling. The measured unit shifts from physical qubits to algorithmic qubits, IBM's term for error-corrected capability.
Google: quality over quantity
Google's roadmap leads with error correction milestones rather than raw qubit counts. Willow showed below-threshold error correction in 2024, meaning adding qubits reduced the error rate, the prerequisite for scale. Google then expanded into neutral atom hardware, hedging across modalities. The implicit promise: a useful error-corrected machine before the end of the decade.
D-Wave: 100 logical qubits by 2032
D-Wave, known for annealers, published a gate-model result in Nature and a roadmap to 100 logical qubits by 2032 using dual-rail erasure qubits. The roadmap matters because it is D-Wave's entry into universal hardware, not its annealing line.
Startups and states
Infleqtion plans a 50 logical qubit machine in Chicago by 2027. Quantinuum already reached 48 logical qubits on Helios, the highest reported count to date, though at a low encoding ratio. State programs like the Illinois Quantum Park fund the facilities these machines need.
The recurring pattern
Every serious roadmap converges on the same shape: physical qubits first, then error correction, then logical qubits, then fault tolerance. The dates differ, the metric does not. A vendor that leads with physical qubit count but has no error correction milestone is behind, whatever the headline number. See Logical Qubits and Fault Tolerance for why the distinction is the whole game.
What works today
The roadmap matters for buyers, but learners and pilot projects do not wait for 2029. Today's machines run real, noisy workloads, and the free tiers give you the same hardware the roadmaps describe. See What Quantum Computers Do in 2026 for the work running now, and getting started to run your first circuit.
- Logical Qubits and Fault Tolerance: the metric every roadmap hides
- IBM's modular cryogenic cells: the Starling building block
- Quantum Processors by Qubit Count: the machines that exist today