IBM’s quantum processors: 3 of 4 promised by year came out on time; the fourth, of 4,000 qubits, did not
- 127 qubits by 2021on time
- 433 qubits by 2022on time
- 1,121 qubits by 2023on time
- 4,000 qubits by 2025not yet
3 of 4 promised by year came out on time; the fourth, of 4,000 qubits, did not.
In 2020 IBM laid out processors of 127, 433 and 1,121 qubits for 2021–2023, and each came out in its promised year. There was no 4,000-qubit processor by 2025: the company switched its goal to error correction.
More
A qubit is a cell of a quantum computer. IBM publishes a «roadmap»: which processor, with how many qubits, in which year. Here are four promised processors and how many years past the deadline each came.
A row is one promise: who, what and by when. The question is whether at least half came true. Answer Yes or No, or just open the answer: each row gets its outcome and what actually came of it, the card a stamp counting those that came true. Quotes and links to the facts are below.
- 4,000 qubits by 2025 — IBM, 2022 years past the deadline
promised by 2025; not happened by 11 October 2026. IBM did not show a 4,000+ qubit processor in 2025: its new processors that year were the 120-qubit Nighthawk and the experimental error-correction chip Loon; the largest remains the 1,121-qubit Condor (2023).
- 127 qubits by 2021 — Jay Gambetta, 2020 years past the deadline
promised by 2021; happened 16 November 2021 (on time). IBM unveiled the 127-qubit Eagle on 16 November 2021 — in the promised year.
IBM Newsroom, «IBM Unveils Breakthrough 127-Qubit Quantum Processor» (16 November 2021)
- 433 qubits by 2022 — Jay Gambetta, 2020 years past the deadline
promised by 2022; happened 9 November 2022 (on time). The 433-qubit Osprey was unveiled on 9 November 2022 — in the promised year.
- 1,121 qubits by 2023 — Jay Gambetta, 2020 years past the deadline
promised by 2023; happened 4 December 2023 (on time). The 1,121-qubit Condor was introduced on 4 December 2023 — in the promised year.
Promises: verbatim from the primary sources, IBM’s roadmaps of 2020 and 2022 (IBM Research blog). The date of the event is from a primary source, linked on the back for each row. The delay is counted from the end of the promised period: «by 2020» from the end of 2020, «in October» from the end of October. Outcome: «came true» — the promised number reached by the deadline («about» within 5%), an event happened before the end of the year promised; «exceeded» — with room to spare; «nearly» — short by no more than 10%; «late» — reached, but after the deadline; «did not come true» — not reached; «came true, then undone» — reached, then reversed. The deadline is the end of the year named.
Qubit count is not a measure of usefulness: without error correction a bigger processor does not necessarily compute better. In 2025 IBM’s new processors were Nighthawk with 120 qubits and the experimental Loon for error correction.
A question for this picture
If the goal changes from «more qubits» to «fewer errors», does the old goal count as missed?