Overview
- IBM announced Nighthawk r3 on Wednesday as a next‑generation processor designed to run 10,000 gate operations across 520 programmable qubits.
- Each r3 chip will contain about 2,034 physical qubit devices and IBM says it will link two r3 processors into a single roughly 1,040‑qubit system in 2027.
- IBM’s current Nighthawk r2 has already delivered large throughput gains with partners, running more than 34 billion circuits and reporting up to 19x speedups and sample rates as high as 100,000 per second.
- The company points to modular demos that underpin the plan, including four l‑couplers that reached about 99.3% state‑transfer fidelity and a two‑module cryogenic assembly that cooled below 15 millikelvin in under five days.
- Reporting shows IBM has adjusted its public scaling approach from linking many small chips toward larger single chips linked in smaller numbers and the company is pairing these hardware steps with tighter integration of classical high‑performance compute to support real workflows and future error‑correction efforts.