D-Wave Advantage2

Overview

D-Wave Advantage2 is a different kind of quantum computer. It does not run circuits of logic gates. Instead it is a quantum annealer: a machine built to find low-energy configurations of a network of interacting qubits, which can represent the best answer to a scheduling, routing, or sampling problem. With more than 4,400 qubits it is the largest superconducting quantum processor by qubit count, though the number is not comparable to the qubit count of a gate-model chip.

D-Wave made the machine generally available in May 2025 through its Leap cloud service, and also sells it as an on-premises system. Customers have included telecoms, manufacturers, logistics companies, and national laboratories.

History and background

D-Wave was founded in 1999 and sold its first commercial machine, the D-Wave One, to Lockheed Martin in 2011. Later generations increased the qubit count and connectivity: the 2,000-qubit 2000Q, the 5,000-qubit Advantage with Pegasus topology in 2020, and then Advantage2. Prototypes of Advantage2 were available on the cloud for several years before the full-scale system launched, and customers ran more than 20 million problems on prototypes.

Advantage2 was announced as generally available on May 20, 2025. Alongside it, D-Wave published a white paper benchmarking the new chip against the previous generation. In 2026 the company bought Quantum Circuits to add gate-model hardware, while continuing to develop annealing for optimization customers.

How it works

Each qubit is a tiny loop of superconducting niobium called a flux qubit, in which current can circulate clockwise, counterclockwise, or in a quantum superposition of both. Couplers between qubits let the user specify how strongly each pair prefers to be aligned or anti-aligned. A programmer turns an optimization problem into such a set of preferences, called an Ising or QUBO model.

During an anneal, the qubits start in a simple quantum state governed by a transverse magnetic field. D-Wave gradually reduces that field and raises the strength of the problem's couplings, so the system evolves toward a low-energy configuration. Quantum tunneling can help the system escape poor local solutions. At the end, the qubits are measured, and each run returns one candidate solution. Users run many anneals and keep the best.

Zephyr's higher connectivity matters because real problems rarely match the hardware's connections one to one. A logical variable that interacts with many others is represented by a chain of physical qubits forced to act together. With 20 neighbors per qubit, chains are shorter and the embedding is more efficient, which improves the quality of answers. D-Wave also provides hybrid solvers that combine the annealer with classical methods to handle problems larger than the chip.

Connectivity

Zephyr topology: each qubit couples to up to 20 others (compared with 15 in the previous Pegasus topology)

Specifications in detail

Key features

• More than 4,400 qubits and more than 40,000 couplers
• Zephyr topology with 20-way qubit connectivity
• 40% higher energy scale, 2x coherence, and 75% lower noise (D-Wave figures)
• Fast-anneal capabilities and hybrid solver integration
• Available through Leap in more than 40 countries and as an on-premises system

Limitations

• It is an annealer and cannot run general-purpose algorithms such as Shor's
• Qubit counts are not comparable to gate-model counts
• Whether annealing beats the best classical optimization methods on commercial problems remains debated
• Results are samples of good solutions rather than guaranteed optima

Timeline

What it is used for

• Workforce and production scheduling
• Network and mobile-tower optimization
• Logistics and routing
• Materials and magnetism simulation, including D-Wave's 2025 simulation paper
• Sampling for machine learning and AI workflows

How it compares

• Compared with gate-model chips like IBM Heron or Google Willow, Advantage2 has many more qubits but is limited to annealing problems.
• Compared with Pasqal Fresnel and QuEra Aquila, it shares a special-purpose, analog flavor, but it uses an annealing protocol on superconducting qubits.
• Compared with NTT's coherent Ising machine, both target Ising-type optimization, with D-Wave using superconducting qubits and NTT using networked optical pulses.

Technical notes

Annealing encodes a problem in the couplings between qubits and relies on slowly changing a magnetic field so the system ends in a low-energy state that represents a good solution. Larger energy scale makes the gap between good and bad solutions bigger relative to noise, and lower noise improves the precision with which problem coefficients are represented. Zephyr's higher connectivity lets many problems be embedded using shorter chains of physical qubits, which also improves solution quality.

Why it matters

D-Wave is the longest-running commercial quantum computing business, and Advantage2 is its most capable machine. Many real business problems are optimization problems, so whether annealing can beat classical solvers on them is one of the most important open questions in applied quantum computing.

It is also a reminder not to compare qubit counts across technologies. 4,400 annealing qubits and 156 gate-model qubits measure different things, and each platform should be judged on the problems it is designed for.

Frequently asked questions

Keep exploring

• Company profile: D-Wave Quantum
• Related processor: Pasqal Fresnel (Orion generation)
• Related processor: NTT 100,512-Spin Coherent Ising Machine
• Related processor: QuEra Aquila
• Related processor: IBM Heron (r1, r2, and r3)

This profile is written for educational purposes and summarizes company announcements, published papers, and vendor specification sheets. Last reviewed: September 30, 2026. Figures marked as company-reported have not all been independently verified, and benchmarks differ between vendors, so compare them with care.

← Back to Hardware Database