Pull to refresh
Logo
What Changed
Following Why Sign Up
Nvidia launches CUDA-Q Logical, a design layer for fault-tolerant quantum computers

Nvidia launches CUDA-Q Logical, a design layer for fault-tolerant quantum computers

New Capabilities

Nvidia's open-source design layer for fault-tolerant quantum computers gains error-mitigation partner Qedma and a validated 1,000-logical-qubit design on silicon-spin hardware.

September 14th, 2026: NVIDIA announces CUDA-Q Logical

Overview

Updated Sep 15

Designing software for fault-tolerant quantum computers has meant rebuilding the toolchain every time a hardware maker changes its qubit type or error-correction code. Nvidia's CUDA-Q Logical, announced Sep 14, gives researchers a single programmable layer to model algorithms, error-correction methods, and processor architectures side-by-side before building hardware.

Early partners are reporting results. Fermilab cut fault-tolerant algorithm development from five months to three weeks, and Infleqtion packed 18 logical qubits into 98 physical qubits. Iceberg Quantum modeled 1,000 logical qubits on Diraq's silicon-spin hardware using 150,000 physical qubits, and error-mitigation startup Qedma has joined the platform.

Why it matters

Whoever owns the design toolchain for fault-tolerant quantum computing shapes which hardware architectures, error-correction codes, and applications actually get built.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

21 days
Fermilab fault-tolerant development time
Down from five months, a 7x speedup using CUDA-Q Logical.
5.4:1
Infleqtion physical-to-logical qubit ratio
Hypergraph-product simplex code uses 98 physical data qubits to encode 18 logical qubits, about a 5x improvement in code rate over surface codes.
150,000
Physical qubits for 1,000 logical qubits
Iceberg Quantum's modeled architecture for Diraq's qubits, roughly 10x fewer than Diraq's previous estimates.

Voices

Curated perspectives — historical figures and your fellow readers.

Sign up to generate historical perspectives on this story.

Log in Sign Up

People Involved

TC
Timothy Costa
Vice president and general manager of quantum at NVIDIA
Oversees CUDA-Q development
 Anna Grassellino
Anna Grassellino
Chief technology officer at Fermilab; director of the Superconducting Quantum Materials and Systems Center
Leading early CUDA-Q Logical adoption
PG
Pranav Gokhale
Chief technology officer and general manager of quantum computing at Infleqtion
Integrated qLDPC library with CUDA-Q Logical

Organizations Involved

Nvidia Corporation
Nvidia Corporation
Semiconductor company
Platform developer

GPU and AI computing giant expanding into quantum software with the open-source CUDA-Q platform.
Fermi National Accelerator Laboratory
Fermi National Accelerator Laboratory
US Department of Energy national laboratory
Early adopter and validation partner

DOE laboratory researching particle physics and superconducting quantum materials.
Infleqtion
Infleqtion
Quantum computing company (neutral-atom)
Partner integrating qLDPC with CUDA-Q Logical

Neutral-atom quantum computing and sensing company listed on the New York Stock Exchange.
Sandia National Laboratories
Sandia National Laboratories
US Department of Energy national laboratory
Developer of QUOPS benchmark

DOE laboratory that developed the QUOPS benchmark now included in NVIDIA CUDA-Q.
Quantum Motion
Quantum Motion
Quantum computing company (silicon spin qubits)
Partner integrating silicon-spin architecture with CUDA-Q Logical

UK startup developing silicon spin qubit processors.
Qedma Quantum Computing
Qedma Quantum Computing
Quantum error-mitigation software company
CUDA-Q integration partner

Quantum error-mitigation company integrating its QESEM software with NVIDIA CUDA-Q.
Iceberg Quantum
Iceberg Quantum
Fault-tolerant quantum architecture developer
Validated Pinnacle qLDPC architecture via CUDA-Q Logical

Architecture developer that modeled 1,000 logical qubits on Diraq's silicon-spin platform.
Diraq
Diraq
Quantum computing company (silicon spin qubits)
Hardware partner in Pinnacle qLDPC validation

Silicon spin-qubit provider whose hardware hosted Iceberg Quantum's Pinnacle validation.

Timeline

6 events Latest: September 14th, 2026 · 2 months ago
Tap a bar to jump to that date

  1. NVIDIA announces CUDA-Q Logical

    Latest Product Launch

    New orchestration layer lets developers model algorithms, error-correction codes, and processor architectures side-by-side. Fermilab reports 7x faster fault-tolerant development; QUOPS benchmark from Sandia debuts in the platform.

  2. Quantum Motion integrates silicon-spin architecture with CUDA-Q Logical

    Partnership

    The UK startup built an automated compilation stack targeting electronic structure calculations, including end-to-end logical resource estimation for the iron-molybdenum cofactor (FeMoco) molecule.

  3. Infleqtion details high-rate error-correction code at IEEE Quantum Week

    Research Result

    Infleqtion presented a hypergraph-product simplex (HGPS) code using 98 physical data qubits to encode 18 logical qubits, about a 5x improvement in code rate over surface codes.

  4. Iceberg Quantum and Diraq validate Pinnacle qLDPC architecture on silicon spin

    Research Result

    Iceberg Quantum compiled its Pinnacle qLDPC error-correction architecture onto Diraq's silicon-spin hardware model using early access to CUDA-Q Logical. The validation confirmed 1,000 logical qubits can run on 150,000 physical qubits, nearly an order of magnitude fewer than surface-code designs.

  5. Qedma integrates QESEM error-mitigation software with CUDA-Q

    Partnership

    Qedma Quantum Computing announced integration of its QESEM error-mitigation software with NVIDIA CUDA-Q. The company, with operations in Toronto and Tel Aviv, works on reducing errors in quantum computations.

  6. CUDA-Q Logical technical paper posted to arXiv

    Technical Publication

    NVIDIA researchers including Alexander McCaskey and Krysta Svore published the technical paper 'CUDA-Q Logical: Retargetable Compilation for Fault-Tolerant Quantum Computing.' It details the compiler's layered architecture from logical programs through quantum error correction microcode to physical gate schedules.

Scenarios

1

CUDA-Q Logical becomes the industry-standard toolchain for fault-tolerant quantum

Possible Resolves by End of 2027

Discussed by: NVIDIA executives and early partners including Fermilab and Infleqtion

NVIDIA's play mirrors its CUDA strategy for GPUs: make the design layer so good that hardware makers have to support it. If IBM Quantum or Google Quantum AI publicly adopts CUDA-Q Logical for production fault-tolerant design work, the platform becomes the de facto standard, and quantum hardware makers compete on qubit quality while NVIDIA controls the software stack above them.

2

CUDA-Q Logical stays within NVIDIA's partner ecosystem

Likely Resolves by End of 2027

Discussed by: Observers noting IBM's and Google's competing investments in their own quantum software stacks

Adoption remains limited to NVIDIA's announced partners: Fermilab, Infleqtion, IQM, Quantum Motion, Diraq, and Sandia. IBM continues pushing Qiskit and Google its own tools, and fault-tolerant design work fragments across proprietary and open-source stacks, with CUDA-Q Logical a strong player but not the standard.

3

Competing open standard for fault-tolerant quantum design emerges

Unlikely Resolves by End of 2027

Discussed by: Qiskit ecosystem developers and the open-source quantum community

A consortium-backed or community-driven toolchain (built on IBM's Qiskit or a new standard) reaches production maturity and is adopted by major hardware makers, splitting the design-tool market and preventing any single vendor from controlling the fault-tolerant quantum software layer.

4

CUDA-Q becomes the common software layer for near-term and fault-tolerant machines

Possible Resolves by Sep 15, 2027

Discussed by: NVIDIA partners including Qedma, which integrated its QESEM error-mitigation software

The Qedma integration extends CUDA-Q from fault-tolerant design into error mitigation for today's noisy machines. If more middleware vendors follow, NVIDIA's platform could serve both near-term and fault-tolerant quantum computers. That would give NVIDIA a hold on the software layer before error-corrected hardware matures.

Historical Context

2 moments from history that rhyme with this story — and how they unfolded.

1980s

EDA tools transform chip design (1980s)

Before electronic design automation (EDA), chip designers drew circuits by hand and simulated them with bespoke scripts. Companies like Cadence and Synopsys automated layout and verification, letting engineers design chips with millions of transistors they never physically touched.

Then

Chip design became faster, cheaper, and reproducible, fueling the microprocessor era.

Now

The EDA industry became a permanent, multi-billion-dollar layer between chipmakers and fabrication, and the same companies still dominate it today.

Why this matters now

CUDA-Q Logical is attempting for quantum hardware what EDA tools did for semiconductors: turn painful, per-architecture manual design into standardized, repeatable tooling. If it works, NVIDIA could occupy the same chokepoint position in quantum that Cadence and Synopsys hold in chips.

2006

NVIDIA's CUDA platform launch (2006)

NVIDIA introduced CUDA, a parallel computing platform that made its graphics processing units programmable for general scientific and engineering workloads. Before CUDA, GPUs were primarily for graphics; researchers had to write assembly-level shader code to use them for computation.

Then

CUDA unlocked GPUs for AI training and HPC, creating a wave of startups and research around GPU computing.

Now

CUDA became the dominant software layer for AI, giving NVIDIA a software moat that competitors have spent years trying to erode.

Why this matters now

CUDA-Q Logical is the same playbook applied to quantum. NVIDIA is betting that whoever owns the software layer above fault-tolerant hardware controls the market, regardless of which qubit technology wins.

Sources

(14)
+8
NVIDIA Newsroom SiliconANGLE The Next Web Quantum Zeitgeist The Quantum Insider Blockchain.News Business Daily Network NVIDIA Developer arXiv Quantum Computing Report Infleqtion Gate News PRNewswire Quantum Computing Report