Google opens $150,000 Kaggle contest on AI-driven quantum circuit compression
Google Quantum AI has launched a $150,000 Kaggle competition to cut qubit and gate counts in key quantum circuits for fusion, optimization and chemistry.
Google Quantum AI has launched a Kaggle competition with a total prize pool of $150,000, aiming to use AI to shrink the quantum circuits needed for practical applications. According to Google’s announcement, the contest focuses on reducing qubit counts and gate counts in benchmark circuits tied to fusion energy, combinatorial optimization and chemistry, while still meeting verification thresholds for accuracy.
What was announced
The competition asks participants to submit improved versions of reference circuits and standalone verifiers released under the Apache 2.0 license. The scoring is based on resource reduction: entrants must pass the verifier for a given task while lowering a cost metric derived from gate and qubit counts.
Google said the effort targets a core constraint in fault-tolerant quantum computing. Even if hardware continues to improve, many useful algorithms remain too resource-intensive. Cutting the size of the circuit lowers not only the direct algorithmic cost, but also the overhead required for quantum error correction.
The prize structure includes $50,000 for first place, $25,000 for second, $10,000 for third, and three $5,000 awards for fourth through sixth. Google also set aside $50,000 for additional prizes, including a $10,000 mid-competition milestone prize and a $10,000 “Longest Leader Prize.” The milestone prize is based on the highest public leaderboard score at 11:59 PM UTC on November 20, and the longest-leader award runs through January 4, 2027.
The technical details
The competition is built around kickmix, an open-source C++20 and Python library inside Google’s Qualtran project. Qualtran’s GitHub repository describes it as a Python library for expressing and analyzing fault-tolerant quantum algorithms.
Google’s companion paper, posted on arXiv, describes the approach as using classically verifiable structure to support AI-assisted compilation. arXiv shows it was submitted on October 16 and revised on October 21.
Kickmix circuits are composed of reversible classical logic, phase operations and measurement-based uncomputation. According to Google, that structure lets a classical computer test candidate circuits efficiently using randomized fuzz testing, rather than fully simulating a general quantum state. The company says this allows approximate correctness to be established with high confidence, though it does not prove exact equivalence.
Google also says kickmix can test circuits at 60 billion gates per second on a single CPU core. That figure refers to conventional verification speed, not quantum hardware execution.
Three benchmark problems
The contest centers on three bottlenecks in larger quantum algorithms.
For fusion energy, the task is to implement a phase proportional to 1/√x from a squared-distance input, a subroutine tied to Coulomb-operator simulations relevant to inertial confinement fusion target design.
For optimization, the benchmark is a Reed-Solomon decoder in superposition, which Google says is the main arithmetic bottleneck in the Decoded Quantum Interferometry algorithm for the Optimal Polynomial Intersection problem.
For chemistry, participants are asked to co-design an accurate factorization and compact loading circuits for parameters in a FeMoCo-related simulation workflow. Google says this benchmark is tied to a recent algorithm for modeling the iron-molybdenum cluster involved in nitrogen fixation.
Why it matters
This is a software-side attempt to bring down the hardware threshold for useful quantum applications. Instead of waiting only for larger processors, Google is trying to automate circuit compression for subroutines that dominate fault-tolerant resource estimates.
Google said internal experiments with its Gemini models have already found improvements beyond circuits reported to date across all three benchmarks. The company published hashes of those internal circuits in its manuscript and said it plans to disclose them after the competition.
What to watch
The key question is whether outside teams can produce materially smaller circuits while staying within the verification bounds. Just as important will be whether Google’s kickmix framework proves broad enough to cover other expensive subroutines in fault-tolerant algorithms, beyond the three benchmarks chosen for this competition.
Sources
Facts on this page come from the sources below, retrieved by DesignQuantum.
- Announcing the verifiable quantum circuit compilation competition on Kaggle opensource.googleblog.com
- Google Launches $150,000 AI Competition to Reduce Quantum Computing Demands thequantuminsider.com
- Leveraging verifiable structure in quantum circuits to find more efficient compilations with agentic AI arxiv.org
- GitHub - quantumlib/Qualtran: Qᴜᴀʟᴛʀᴀɴ is a Python library for expressing and analyzing Fault Tolerant Quantum algorithms. github.com
- Qualtran/kickmix at main · quantumlib/Qualtran github.com