Within nineteen days in June and July 2026, two research teams—one simulating the quantum dynamics of charged black holes, the other accelerating the search for stationary points of any smooth function—quietly redrew the boundary between problems a quantum algorithm can solve efficiently and those classical machines must own. Both papers stretch the concept of quantum advantage in directions that have nothing to do with factoring large integers, and both arrive as quantum hardware crosses long-promised performance thresholds. [arXiv:10.1140/epjc/s10052-026-16055-7]
This matters because the two signals expose the same underlying trend: quantum algorithm research has fragmented into a portfolio of approaches, each targeting a different class of problem. The Reissner-Nordström paper refines a decades-old quantum gravity framework using affine variables and variational circuit-style ansätze. The Li et al. stationary-point algorithm delivers a polynomial quantum speedup for non-convex optimization via quantum gradient estimation. Neither depends on the other, and yet both demonstrate that quantum software is becoming useful long before fault-tolerant hardware arrives.
How It Works
The black-hole paper takes a quantum-gravity approach. Its authors apply affine quantization—a technique well-suited to positive-definite geometric variables—to the charged Reissner-Nordström spacetime, the standard solution for a non-rotating black hole carrying electric charge. They reduce the problem to a minisuperspace model, a simplified arena with just enough degrees of freedom to capture the essential physics.
"The resulting Wheeler-DeWitt equation becomes separable, yielding Hermite-polynomial modes in one sector and Gaussian-like radial solutions in the other."
That separability is the breakthrough: it lets physicists construct normalizable semiclassical wave packets and compute probability distributions in minisuperspace that earlier canonical quantization methods could not handle. The Li et al. algorithm tackles a different beast. Finding ε-stationary points of a twice-differentiable function—locations where the gradient's magnitude falls below ε—is a workhorse problem in machine learning, statistics, and operations research. Classical comparison-based methods need Õ(n²/ε) oracle queries, while the new quantum approach, by evaluating the function in superposition and using quantum phase estimation to extract gradient components, cuts this to Õ(n/ε^1.5).
For n=10,000 variables, that is the difference between roughly 10⁸ and 3×10⁵ oracle calls—a thousand-fold reduction in a routine that runs every day inside every large model. The mathematical connection is subtle but real. The Hermite polynomials of the gravity paper and the gradient-estimation subroutines of the optimization paper both exploit properties of continuous-variable quantum systems, the same toolkit now driving Topological Qubits designs at Microsoft and the neutral-atom hardware at QuEra and Pasqal. The lineage runs through Seth Lloyd of MIT, whose 1996 proof established that quantum systems can simulate other quantum systems efficiently, and through John Preskill of Caltech, who in 2018 coined the term NISQ for the noisy intermediate-scale era that defines deployment in 2026.
Who's Moving
Hardware is the limiting reagent, and the vendors are public. International Business Machines (NYSE: IBM) operates the largest installed base, anchored by the 1,121-qubit Condor processor and the 156-qubit Heron, backed by a $3.5 billion quantum research commitment announced in 2024. Alphabet (NASDAQ: GOOGL) leads on the error-correction front; its 105-qubit Willow chip, unveiled in December 2024, was the first device to demonstrate below-threshold quantum error correction, the threshold below which adding qubits reduces logical error rates. IonQ (NYSE: IONQ) and Quantinuum are pursuing trapped-ion architectures, while Rigetti (NASDAQ: RGTI) stays with superconducting circuits, and PsiQuantum bets on photonic qubits backed by more than $940 million in private capital.
Each vendor is also the largest customer of the quantum software stack emerging around it: variational circuit libraries, hybrid quantum-classical compilers, and noise-aware transpilers are now separate product lines inside the major clouds. Ibm Quantum Platform and Google Quantum Ai each host tens of thousands of registered algorithm developers, while Rigetti's Quil and IonQ's toolchains serve smaller but growing communities. The competitive pressure is shifting from qubit count alone to the depth of the surrounding software ecosystem. Hardware without algorithms, the industry has learned the hard way, is a Ferrari without a steering wheel.
Why 2026 Is Different
Three timelines now run in parallel. Within 12 months, NISQ devices—those with 50 to a few thousand noisy qubits—run variational circuit prototypes for materials science and small optimization instances on hardware that exceeds the threshold for useful simulation. Within 3 years, the first fault-tolerant logical qubits with circuit depths exceeding 10⁶ operations arrive at IBM, Google, and Quantinuum, enabling end-to-end runs of algorithms like Li et al.'s on problems of practical size. Within 5 years, McKinsey's projected $1 trillion in cumulative quantum value creation by 2035 begins to materialize in optimization, cryptography, and molecular simulation.
The lesson is structural. Quantum algorithms are no longer a single bet on Shor's factoring routine; they are a portfolio spanning black holes, smooth functions, materials, and machine learning, with affine quantization and quantum gradient estimation as parallel tools of the trade. The 19-day window between the two papers is the smallest unit of a much larger shift that started with Peter Shor's 1994 paper and accelerated through Google's 2019 sampling demonstration. In short: a single quantum algorithm now finds the stationary points of any n-variable function in Õ(n/ε^1.5) oracle queries—a 1,000× speedup over classical methods at practical scales.
Frequently Asked Questions
What is a quantum algorithm?
A quantum algorithm is a step-by-step procedure that runs on a quantum computer and exploits superposition, entanglement, and interference to solve specific problems faster than any known classical method. Unlike classical bits that hold a single value, qubits can exist in superpositions of 0 and 1, allowing the algorithm to evaluate many inputs simultaneously. Famous examples include Shor's 1994 factoring algorithm and Grover's 1996 search algorithm. Modern quantum algorithms are increasingly built as hybrid quantum-classical routines, where a classical optimizer trains a variational circuit.
How does the Li et al. quantum algorithm compare to classical methods for finding stationary points?
The new approach finds an ε-stationary point of a twice-differentiable function in Õ(n/ε^1.5) oracle queries, down from Õ(n²/ε) for comparison-based classical algorithms such as variance-reduced gradient descent. The technique replaces multiple classical function evaluations with a single quantum query that estimates gradient components in superposition. For problems with up to 10,000 variables, the speedup is roughly 1,000×, an advantage that holds under both arithmetic and black-box oracle models.
When will quantum algorithms achieve commercial advantage?
Quantum advantage on contrived sampling problems was first demonstrated in 2019 by Google. Commercial advantage—where a quantum algorithm runs faster or cheaper than the best classical alternative on a problem somebody pays to solve—arrives between 2029 and 2031. IBM, Google, and Quantinuum have public roadmaps targeting 200 logical qubits and circuit depths of 100 million gates by 2029. The first commercial deployments land in molecular simulation for pharmaceuticals and in portfolio optimization for finance, where non-convex problems with thousands of variables are routine.
Which companies are leading in quantum algorithm development?
IBM (NYSE: IBM) leads in installed base and hybrid quantum-classical software through its Qiskit platform. Alphabet (NASDAQ: GOOGL) leads in error-correction milestones and is investing heavily in algorithm libraries for chemistry and optimization. IonQ (NYSE: IONQ), Rigetti (NASDAQ: RGTI), and Quantinuum are pursuing distinct hardware paths. PsiQuantum is building photonic systems backed by more than $940 million in private capital. Smaller players such as Atom Computing, QuEra, and Pasqal focus on neutral-atom hardware and co-designed algorithms.
What are the biggest obstacles to quantum algorithm adoption?
Three obstacles dominate. First, current NISQ devices lack the qubit counts and error rates needed for fault-tolerant execution of large quantum algorithms. Second, end-to-end quantum speedups on commercially valuable workloads have not yet been demonstrated outside contrived benchmarks. Third, the quantum software stack—compilers, error-correction decoders, calibration tools—still requires specialized expertise, raising the cost of deployment. Hybrid quantum-classical workflows and improved error-correction codes are the most promising mitigations, but each is years from maturity.