Rigetti Launches 108-Qubit Cepheus System, the Largest Modular Chiplet-Based Quantum Computer Available via the Cloud
Rigetti Computing has made its 108-qubit Cepheus-1-108Q system generally available through its own cloud platform and Amazon Braket, marking the first gate-based quantum processor exceeding 100 qubits on AWS.
Overview
Rigetti Computing announced the general availability of Cepheus-1-108Q, a 108-qubit superconducting quantum processor built from twelve interconnected 9-qubit chiplets, according to the company’s official press release. The system is now accessible to customers through Rigetti’s Quantum Cloud Services platform and through Amazon Braket, making it the first gate-based quantum device exceeding 100 qubits available on AWS’s quantum service, as confirmed by Amazon Web Services.
The launch triples the qubit count and chiplet count of Rigetti’s previous 36-qubit Cepheus-1-36Q system and represents the industry’s largest modular quantum computing system to date.
What We Know
Cepheus-1-108Q is organized as a 3x4 array of twelve 9-qubit chiplets with tunable couplers and intermodule couplers connecting the individual chips, according to Amazon’s announcement. This modular chiplet-based architecture is central to Rigetti’s scaling strategy, allowing the company to manage yield and fabrication complexity by tiling smaller, well-characterized processor units rather than manufacturing a single monolithic chip.
The system achieves a median two-qubit gate fidelity of 99.1 percent at a gate speed of approximately 60 nanoseconds, along with a median single-qubit gate fidelity of 99.9 percent, according to Rigetti’s press release. On prototype hardware, Rigetti has demonstrated two-qubit gate fidelity as high as 99.9 percent at 28 nanoseconds using an adiabatic controlled-Z (CZ) gate scheme. The company expects the production system to reach a median 99.5 percent two-qubit gate fidelity later in 2026.
A notable technical change in this generation is the shift from iSWAP gates, used on previous Rigetti processors, to CZ gates. According to AWS, CZ gates provide higher resilience to phase errors common in superconducting systems, and Rigetti’s adiabatic CZ implementation further reduces leakage errors, enabling deeper circuits for use cases such as chemical simulation, combinatorial optimization, and machine learning.
Several hardware and fabrication improvements underpin the performance gains. Rigetti’s Alternating-Bias Assisted Annealing fabrication technique improves qubit frequency targeting and reduces defects, while upgraded control electronics deliver improved signal-to-noise ratios for qubit readout, as reported by The Quantum Insider. Superconducting qubit gate speeds operate approximately 1,000 to 10,000 times faster than those of trapped-ion and neutral-atom systems, a point Rigetti emphasizes as a competitive advantage.
Dr. Subodh Kulkarni, Rigetti’s CEO, stated that the system “validates our ambitious approach to scaling quantum computers” and that the chiplet-based architecture is “paving the way toward higher fidelity, higher qubit systems” that will enable fault-tolerant quantum computing, according to the press release. Eric Kessler of Amazon Braket noted that “Cepheus-1-108Q delivers improved fidelities allowing customers to push to wider and deeper circuits,” as reported by The Quantum Insider.
Users can develop quantum programs for the system via the Braket SDK, Qiskit, CUDA-Q, and Pennylane, with pulse-level control available for researchers requiring low-level hardware access for noise studies and error mitigation development, according to AWS.
What We Don’t Know
Rigetti has not published detailed benchmarks comparing Cepheus-1-108Q directly against competing systems from IBM, Google, IonQ, or Quantinuum. The company’s two-qubit gate fidelity of 99.1 percent, while a significant improvement over its prior generation, trails the highest fidelities reported by some trapped-ion and neutral-atom platforms, which have demonstrated two-qubit gate fidelities above 99.5 percent. Whether Rigetti’s speed advantage compensates for this fidelity gap in practical applications remains an open question.
The company has acknowledged that coherence time is currently the primary factor limiting further performance improvements, but has not disclosed specific T1 or T2 coherence figures for the production system. The timeline and technical milestones required to reach the company’s stated goal of quantum advantage within approximately three years have also not been detailed beyond the near-term fidelity target of 99.5 percent.
It is also unclear how inter-chiplet connectivity and crosstalk affect algorithm performance at scale, a challenge inherent to any modular quantum architecture that tiles multiple smaller processors together.
Analysis
Rigetti’s chiplet approach addresses one of the most stubborn engineering constraints in superconducting quantum computing: fabrication yield. Monolithic processors require every qubit on a single chip to meet specification simultaneously, and defect rates rise with chip area. By constructing a processor from twelve smaller, individually testable chiplets, Rigetti can screen for defects at the module level and discard only the faulty units rather than an entire wafer-scale processor. This manufacturing strategy becomes increasingly important as the industry targets systems with hundreds or thousands of qubits.
The shift from iSWAP to adiabatic CZ gates is significant beyond raw fidelity numbers. CZ gates are natively compatible with the error correction codes most widely studied for fault-tolerant quantum computing, including the surface code, potentially reducing the overhead needed to translate between native gate sets and the operations required by error correction protocols.
The availability of the system through Amazon Braket lowers the barrier to experimentation for enterprise and academic users who may not maintain dedicated quantum hardware access. As quantum processors grow more capable, cloud-based access is likely to remain the primary channel through which most users interact with the technology, making platform integrations a competitive differentiator alongside raw hardware performance.