On July 27, 2026, AT&T and D-Wave Quantum announced a contract to expand the use of D-Wave's quantum computing technology across AT&T's network operations. In an initial use case, the companies say a certain network optimization process was shortened from about one hour to under 15 seconds. Taking the disclosed figures at face value, that's at least a 240-fold difference in processing time.
However, what got faster is neither a communications line nor fault recovery—it's a single computational process. The scale of the problem, the classical computing environment used for comparison, and the quality of the resulting solution have not been disclosed. What this contract sets in motion is a phase of verifying whether quantum annealing, connected to AT&T's existing agentic AI, can be used for fault response and infrastructure planning. How should carriers evaluate this figure of 15 seconds in practical terms?
What the "240-fold" Figure Actually Measures
One hour equals 3,600 seconds, so the difference against under 15 seconds exceeds 240-fold. That's as far as the performance figures in the joint announcement go. The target is described only as a "network optimization workload"—the number of variables and constraints, the objective function, which solver was used for comparison, the CPU/GPU configuration, and the number of trials remain unknown. Nor is it clear whether the processing time includes data preparation and result verification.
In optimization, value isn't determined by speed alone. A solution obtained quickly needs to have an objective function value equal to or better than that of conventional methods, and it must satisfy operational constraints. For the under-15-second result, there's no information on the gap from the optimal solution, reproducibility, or a comparison against what a purely classical computation would produce given the same amount of time. This figure therefore demonstrates a large time difference in the initial use case, but it does not prove "quantum advantage" over classical computing.
The use cases AT&T plans to investigate going forward are candidates: fault detection and response, technician routing, network construction planning, and traffic management. The announcement does not state that all four areas have entered commercial operation. Whether the reduction in processing time actually improves recovery time, travel distance, capital expenditure, or communication quality remains to be measured.
What Annealing Speeds Up Is Search, Not Communication
Quantum annealing is a computational method for searching among many candidates to find combinations that meet an objective. According to D-Wave's explanation, problems like delivery routing or scheduling are converted into energy minimization problems, and a quantum processor returns low-energy solutions. In network operations, this corresponds to decisions such as who to dispatch where, which equipment to add, and how to allocate limited resources.
In addition to sending problems directly to the quantum processor, D-Wave also offers hybrid solvers that combine classical and quantum computing. According to official documentation, in this configuration the classical side selects the difficult parts of a problem to hand off to the quantum processor, or splits a large problem and recombines the results. This makes it easier to handle problems of practical scale, but it also means the processing time includes the work done by classical algorithms.
It's not possible to tell from the announcement whether AT&T's initial use case employed the quantum processor directly or used a hybrid solver. Since neither a product name nor a solver name is given, we also can't determine how many of the 15 seconds were spent on quantum processing, or how much that portion contributed to the overall time difference. Communication packets don't pass through a quantum computer; rather, the computation that generates operational options runs externally.
Where Exactly Does Quantum Computing Fit Into Agentic AI?
AT&T's policy is to incorporate quantum annealing into the existing tools underpinning its agentic AI solutions. In a network fault-response scenario the company described in November 2025, an AI agent narrows down the problem location from telemetry, cross-references change history and known faults, and creates a ticket. A separate agent proposes a fix, with human technicians overseeing the judgment calls along the way.
Where an optimization solver could fit in is the part that computes "which response to choose" after the situation has been assessed. If technician visit sequences or equipment plans could be recalculated repeatedly, agents wouldn't need to rely on outdated plans for long. But AT&T has yet to disclose the recalculation cycle in production environments, or the results of connecting the 15-second process to actual decision-making.
The joint announcement also states that existing agentic tools reduced customer downtime by 12 million hours in 2025. This figure reflects the performance of existing tools—it is not the effect of introducing D-Wave. The number of customers covered and the calculation method are also undisclosed. In future comparisons, unless the case with existing tools alone is separated from the case with quantum computing added, the value of the additional computing resource cannot be measured.
Heavier Than a 5-Minute Solve: NTT DOCOMO's Commercial Network Data
When it comes to introducing quantum annealing into telecom networks, NTT DOCOMO has published measurement results that are a step further along. In February 2026, the company deployed a "TA-List optimization algorithm" into its commercial network to optimize tracking areas, which group together base stations, and the higher-level groupings above them. It converts into QUBO form a problem that simultaneously reduces location registration signals sent by devices when moving and paging signals sent by base stations upon incoming calls, and solves it in about 5 minutes.
DOCOMO reduced location registration signals by 65.3% and average base-station paging signals by 7.0% during peak hours on a given day in a specific area. Its predecessor method from 2024 also reduced peak-hour paging signals by up to about 15%, and average paging signals by about 7%, across the Tokai, Chugoku, and Kyushu areas. The company said this corresponds to roughly 1.2 times more capacity for connecting devices during periods when incoming calls are concentrated, and began rolling it out sequentially to base stations nationwide starting in July of that year.
Since DOCOMO and AT&T are solving different problems, lining up 5 minutes against 15 seconds doesn't settle which technology is superior. What the comparison reveals is the evidence needed beyond solve time. DOCOMO reflected its optimization results in a commercial network and measured exactly how much control signaling was reduced. Only once AT&T also presents operational metrics matching its chosen use cases—such as fault recovery time or technician travel distance—can the 240-fold-plus time difference be connected to customer value.
Gate-Based Evaluation Is on a Different Timeline
AT&T also says it will evaluate D-Wave's future gate-model quantum computers, exploring applications in quantum security and quantum communications. This is separate from the quantum annealing use case for which processing times have already been given. It does not mean post-quantum cryptography has been deployed, nor that a quantum communications network is being built.
The gate-model roadmap D-Wave presented at its Investor Day in June 2026 sets a goal of a system capable of executing over 1 million gate operations on 100 logical qubits by 2032. AT&T's announcement explicitly refers to evaluating "systems to be offered in the future," so this future target cannot be added to current network operational capability.
The value of this contract will be determined not by the eye-catching figure of 15 seconds, but by what gets disclosed afterward. If AT&T aligns problem scale, classical comparison conditions, and solution quality, and manages to bring any of the four use cases into commercial operation to improve recovery time or planning accuracy, quantum annealing will establish itself not as a device that makes networks faster, but as a tool that updates the decisions that run them.