In February 2021, Winter Storm Uri triggered a cascading failure across Texas—frozen natural gas supplies and halted wind turbines combined to knock out roughly half the state's generating capacity, some 46,000 megawatts, all at once. This wasn't a single equipment failure; it was hundreds of failures cascading like an avalanche within a matter of hours. The blackout lasted four days and claimed 246 lives.
The North American Electric Reliability Corporation (NERC), which sets reliability standards for the North American grid, mandates an "N-2 standard" for transmission systems. This requires that the system maintain stable operation even after two sequential failures occur. In N-2, "N" refers to the total number of components making up the grid (transmission lines, transformers, generators, etc.), and "2" refers to the number of failures being considered.
When this standard was established, threats to the grid were mainly single equipment failures. But by the 2020s, the nature of the threat had changed. Wildfires and hurricanes intensified by climate change, ransomware and state-sponsored cyberattacks, physical sabotage—these threats don't necessarily occur in isolation. A compound scenario in which a wildfire burns transmission lines while a simultaneous cyberattack targets control systems is not merely theoretical.
As Sid Suryanarayanan, Senior Chief Engineer at Eaton, stated in the official announcement: "Extreme weather, wildfires, and physical and cyber threats are creating unprecedented risks to power reliability and security, and we need tools that can consider multiple simultaneous failures."
When 4 Out of 300 Pieces of Equipment Fail Simultaneously, 330 Million Calculations Emerge
The fundamental problem with the N-2 standard lies in computational complexity.
If a grid has $N$ components and $k$ components fail simultaneously, the number of failure combinations to examine is given by the binomial coefficient . In plain terms, this calculates "how many ways there are to select $k$ items simultaneously from $N$ items."
Let's look at concrete numbers. For a transmission system with 300 components, analyzing four simultaneous failures (N-4) yields approximately 330,790,000 combinations. For each combination, a power flow calculation—solving nonlinear equations to determine how much current flows through which lines—must be performed. Even assuming each calculation takes just one second, covering all patterns would take more than 10 years.
| Analysis Standard | Number of Combinations (118-bus system) | Total Computation Time at 1 Second Each |
|---|---|---|
| N-1 (single failure) | 166 | About 3 minutes |
| N-3 (3 simultaneous failures) | Over 790,000 | Over 9 days |
| N-4 (300-component system) | About 330,790,000 | About 10.5 years |
Research published on arXiv (2310.04213) estimates that while N-1 analysis on the IEEE 118-bus test system takes 3 minutes, N-3 analysis requires more than 9 days on a single computer. This is just for a small-scale test system—the actual North American grid spans approximately 527,000 miles of transmission lines at 100 kV or above (per NERC's 2025 Reliability Report), with about 5,832 transformers and 30,014 transmission circuits. N-4 analysis on the real grid is effectively impossible at this scale using classical computers.
The Structural Reason the US Air Force Chose Quantum Computing
On August 6, 2026, power management giant Eaton (2025 revenue of $27.4 billion, operating in 180 countries) announced it had secured a $7 million, 24-month contract from AFRL. Its partners are Infleqtion, which develops neutral-atom quantum computers, and Pennsylvania State University, which handles machine learning and artificial intelligence.
The reason AFRL is focusing on quantum computing lies in the structure of this problem: it is a combinatorial optimization problem that is NP-hard (solvable in polynomial time on a non-deterministic Turing machine, but requiring exponential time in the worst case on classical computers). Quantum computers offer a theoretical advantage by using the superposition and interference of qubits to search vast candidate spaces in parallel.
However, current quantum computers remain in the "Noisy Intermediate-Scale Quantum (NISQ)" era, and demonstrated cases of "quantum advantage"—where quantum computers solve large-scale problems faster than classical computers on real-world problems—remain limited. The project's adoption of a "hybrid quantum-classical" approach reflects this reality. It divides the work between what quantum computers excel at (searching combinatorial spaces) and what classical computers excel at (precise power flow calculations, constraint verification).
The hardware provided by Infleqtion arranges neutral atoms (rubidium and cesium) using optical tweezers and performs quantum gate operations through Rydberg excitation. The company's Sqale platform features over 100 physical qubits and has achieved a two-qubit gate fidelity of 99.73%. The company has published a roadmap targeting 30 logical qubits (with error detection) by 2026, 100 logical qubits by 2028, and 1,000 logical qubits by 2030.
What the Project Will Actually Do
Over the 24-month research period, the team will pursue four main tasks.
First, developing new quantum algorithms. This involves designing methods to convert grid vulnerability searches into a form that quantum circuits can process (encoding the problem as a Hamiltonian, i.e., an energy function). Prior research has applied QAOA (Quantum Approximate Optimization Algorithm) to this type of problem, with validation performed on IEEE 9-bus and 14-bus systems. However, one report notes that applying QAOA to N-k analysis on a 14-bus system required about 48 hours on a classical computer, making the limitations of current hardware clear.
Second, optimizing circuits for hybrid quantum-classical computation. This requires circuit designs that minimize circuit depth (the number of gate operation stages) while accurately representing the problem's structure.
Third, experimentation across multiple quantum hardware platforms. In addition to Infleqtion's neutral-atom approach, the team will validate using other approaches (superconducting, ion trap, etc.) to evaluate the algorithm's hardware dependency.
Fourth, error mitigation and correction. Current quantum computers are noisy, and calculation results contain errors. How to suppress this noise and determine how reliable the resulting answers can be is a bottleneck for practical application.
The final deliverable is a proof-of-concept demonstration that combines existing quantum hardware, new algorithms, and machine learning to address real-world grid challenges. The plan is to demonstrate how utility operators could visualize complex grid states, evaluate multiple threat scenarios, and accelerate decision-making during emergencies.
The Competitive Landscape of Quantum Computing and the Power Grid
The Eaton-AFRL project is just one part of a broader movement to apply quantum computing to the power grid.
Infleqtion itself has received $6.2 million from the US Department of Energy's ARPA-E (Advanced Research Projects Agency-Energy) for its ENCODE (Enhancing Neutral-atom Computers for Optimizing Delivery of Energy) project, working with Argonne National Laboratory, EPRI, and ComEd to advance grid optimization through quantum computation. This is ARPA-E's first quantum energy project, focusing on grid "delivery optimization."
In July 2025, IonQ and Oak Ridge National Laboratory announced results using a hybrid approach with a 36-qubit ion-trap quantum computer to solve the Unit Commitment problem—determining generator startup and shutdown schedules. The experiment optimized schedules for 26 generators over 24 hours using a quantum-classical hybrid approach, as part of DOE's GRID-Q project.
In March 2026, DOE also launched a quantum algorithm competition co-hosted with Quantum World Congress. This is a broad call for algorithms that solve grid planning challenges using quantum and hybrid computation, with experts from national laboratories serving as judges.
| Project | Funding | Quantum Approach | Target Problem | Timeline |
|---|---|---|---|---|
| Eaton/AFRL (this project) | $7 million | Neutral atom (Infleqtion) + other approaches | N-k simultaneous failure vulnerability analysis | Aug 2026–2028 |
| Infleqtion/ARPA-E ENCODE | $6.2 million | Neutral atom | Grid delivery optimization | 2025– |
| IonQ/ORNL GRID-Q | Undisclosed (separate EPB deal worth $22 million) | Ion trap | Unit Commitment | Results announced July 2025 |
| DOE Quantum Algorithm Competition | Undisclosed | Method-agnostic | Grid planning in general | Started March 2026 |
Three Unresolved Problems Facing the Demo Two Years From Now
The success of this project hinges on whether, two years from now, the proof-of-concept demo can "solve a practically meaningful problem at a meaningful scale, with meaningful accuracy and speed compared to classical computing." At least three challenges stand in the way.
First, the qubit count wall. Encoding N-4 analysis for a real-scale grid (thousands to tens of thousands of components) into quantum circuits requires orders of magnitude more than the current 100 physical qubits. Even if 30 logical qubits are achieved in 2026 as Infleqtion's roadmap projects, the scale of problems that can be handled will likely still remain at the test-system level.
Second, QAOA's limited expressive power. Prior research has reported that shallow-depth QAOA circuits fail to adequately explore the solution space for large-scale problems, leaving errors of about 7% from the optimal solution. Deepening the circuit improves accuracy, but noise increases proportionally. How to break through this trade-off is at the heart of algorithm design.
Third, the dynamic nature of cyber threats. While physical equipment failures can be modeled relatively statically, cyberattacks represent a dynamic problem in which adversaries adapt. How should the possibility that attackers learn and evade the defenders' analytical methods be handled? Eaton's announcement offers no specific approach on this point.
According to NERC's cybersecurity incident reporting (CIP-008-6), a total of 17 cybersecurity events were reported over the five years from 2021 to 2025, none of which led to large-scale blackouts. However, the 2024 report notes "increased sophistication" of attacks, and NERC itself acknowledges that the low number of reported incidents may not reflect minimal damage but rather ambiguity in reporting standards. With the true scale of the threat difficult to grasp, is it enough to simply make the defenders' tools more sophisticated? This question remains—on a dimension separate from quantum computing performance.