TL;DR
Get business pricing on networking and server gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Alice & Bob demonstrated GPU decoding of synthetic measurement data for its LDPC-cat error-correction codes using NVIDIA NVQLink and Quantum Machines hardware. The proof of concept processed multiple sets of measurements in parallel, but did not achieve real-time decoding; the company says it will work on the data pipeline next.
Alice & Bob has demonstrated GPU decoding of synthetic measurement data for its LDPC-cat error-correction codes using NVIDIA’s NVQLink architecture and Quantum Machines control hardware. The proof of concept showed improved decoding efficiency when multiple sets of measurements were processed in parallel, but did not achieve real-time decoding, a requirement for keeping pace with a running quantum computer.
The company said researchers tested whether measurement data could be transferred from control electronics to a GPU through NVQLink. The setup used Quantum Machines’ OPX1000 and Open Acceleration Stack, a framework that connects control hardware with GPUs and other classical processors. The work used synthetic data, rather than reporting a demonstration on measurements generated in real time by a working quantum processor.
Alice & Bob described the result as a proof of concept for decoding LDPC-cat codes on GPUs through NVQLink. It reported that decoding became more efficient when several sets of error measurements, or syndromes, were handled in parallel. The company did not provide a measured real-time decoding result in the supplied announcement, and said it now plans to optimize the pipeline that moves and processes measurement data.
NVQLink is NVIDIA’s open system architecture for connecting GPU computing with quantum processors. It is designed to support workloads that need communication between quantum and classical hardware, including error-correction decoding, calibration and hybrid quantum-classical applications. Alice & Bob says it is working with NVIDIA on the next stage of the effort.
Scaling Decoding for Cat Qubits
Quantum computers accumulate errors, and error correction depends on classical systems interpreting measurements quickly enough to help correct them. The demonstration matters because it tests one part of that classical workload: whether GPU resources can decode Alice & Bob’s code using a link designed for communication between quantum-control hardware and accelerated computing.
Parallel processing may offer a way to expand decoding capacity as a system handles more measurements. But efficiency when processing batches is not the same as meeting the timing demands of a live quantum processor. Until the data-transfer and processing pipeline can keep pace, this result is evidence of technical feasibility, not proof of real-time fault-tolerant operation.
The work also connects a specific error-correction approach to an existing hardware and software stack. If later tests establish the needed speed and reliability, GPU-based decoding could help researchers scale classical processing alongside cat-qubit systems. The announcement does not establish that this performance has been reached.
GPU server for quantum error correction
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier Alice & Bob GPU Work
Alice & Bob developed LDPC-cat codes for its cat-qubit architecture. The company introduced the codes in a paper, “LDPC-cat codes for low-overhead quantum computing in 2D,” published in January 2025. The October demonstration applies GPU decoding to that code family through NVQLink, but the reported test was limited to synthetic data.
The companies have announced other GPU-related work. In June 2025, they said NVIDIA CUDA-Q had been integrated into Dynamiqs, Alice & Bob’s quantum processor simulation library. In March 2026, they reported a runtime of one hour and 57 minutes for decoding simulated syndrome data for Alice & Bob’s Elevator Codes using CUDA-Q. That earlier result involved a different code and should not be treated as a real-time result for LDPC-cat decoding.
““Alice & Bob’s work shows how NVQLink can in fact accelerate the real-time decoding needed to turn cat qubits into fault tolerant quantum-GPU supercomputers.””
— Sam Stanwyck, NVIDIA director of quantum product
As an affiliate, we earn on qualifying purchases.
Real-Time Performance Still Unreported
The announcement does not provide a measured decoding latency, throughput figure, or comparison against the timing requirements of Alice & Bob’s quantum hardware. It also does not report a test using live measurement data from a quantum processor. As a result, how close the pipeline is to real-time operation remains unclear.
The company says several measurement sets could be decoded more efficiently in parallel, but it gives no quantitative measure of that improvement in the supplied material. The reliability of the result under operating conditions, and whether performance scales as the amount of data increases, are also not established here.
high performance GPU for parallel computing
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Pipeline Optimization Comes Next
Alice & Bob says it will optimize the pipeline carrying measurement data through NVQLink to the GPU, working with NVIDIA. The stated objective is to reach the performance needed for real-time decoding. The companies have not announced a deadline, a next public test, or a target latency in the supplied material.
Further updates would need to clarify whether the revised system can process live quantum measurements, how quickly it can return decoding results, and whether that speed holds as workloads grow. Until those details are reported, the October result remains a synthetic-data proof of concept.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did Alice & Bob demonstrate?
Alice & Bob demonstrated GPU decoding of synthetic measurement data for its LDPC-cat codes using NVIDIA NVQLink and Quantum Machines control hardware. The test also examined processing several measurement sets in parallel.
Did the demonstration achieve real-time decoding?
No. Alice & Bob said the proof of concept did not constitute real-time decoding. The company identified pipeline optimization as its next step toward keeping pace with its quantum computers.
What is NVIDIA NVQLink?
NVQLink is NVIDIA’s open architecture for connecting GPU computing with quantum processors. It is intended for tasks that require communication between quantum and classical hardware, including error correction.
What hardware was used in the test?
The assessment used Quantum Machines’ OPX1000 and Open Acceleration Stack to connect control electronics with GPUs and other classical processors.
What remains to be shown?
The announcement does not give a latency or throughput result, or demonstrate decoding live measurements from a quantum processor. Whether the system can meet real-time requirements at larger workloads has not been established.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
