#quantinuum

Quantum enters large-scale logical computing: hybrid HPC integration and error correction innovations

Quantinuum News

Quantinuum has announced the transition from experimental quantum hardware to large-scale logical computing with multiple milestones. By achieving “beyond break-even” performance with almost 100 logical qubits and complex hybrid workflows with the world's most powerful supercomputers, the company overcame years-old hurdles.

“Skinny Logic” Breakthrough The latest quantum error correction (QEC) study is the company's biggest achievement. Quantum mechanics researchers have devised a highly effective method for building logical qubits, which are entangled physical qubits that exchange information to shelter it from noise. With the 98-qubit Helios quantum processor, researchers found 64 error-detected and 48 error-corrected logical qubits.

“Skinny Logic” uses high-rate encoding to “cut overhead to the bone.” The “no free lunch” rule of quantum physics states that high-quality logical qubits require a lot of physical qubits. The researchers achieved a world-record 2:1 physical-to-logical ratio for error-corrected qubits by “nesting” super-efficient quantum error-detecting codes, including the famed “iceberg code”.

The logical qubits outperformed the physical ones in every test, typically by a factor of 10 to 100. The business's “holy grail” is “beyond break-even” fidelity, which proves data encoding improves computing reliability over bare hardware.

Representation of Reality in 3D Quantinuum demonstrated the applicability of logical qubits by simulating quantum magnetism on a wide scale. Traditional studies use 1D or 2D models to speed computations, but the Helios processor's “all-to-all” connectivity allowed researchers to mimic 3D material interactions. This was possible due to the trapped-ion design's mobile qubits that can communicate across the CPU.

The group created a 94-logical qubit GHZ state, or “cat” state, with 94.9% fidelity to demonstrate system-wide entanglement. This test shows Helios hardware's ability to sustain complicated entanglement beyond approximately 100 logical units, breaking previous un-encoded marks.

Quantum-HPC Hybrid Quantinuum demonstrated that quantum computers can be integrated into HPC ecosystems beyond pure quantum accomplishments. In conjunction with RIKEN in Japan, the world's most powerful supercomputer, Fugaku, was coupled with the Reimei quantum computer.

For the first time, a thorough scientific methodology was used to study chemical processes in proteins at various structures. Reimei modeled the sophisticated quantum mechanics of the molecule's "active site" while Fugaku computed baseline electronic structure. Modern quantum technologies can improve classical systems for materials and medicinal research via a hybrid strategy.

AI-Powered NVIDIA Integration and Exploration AI-driven algorithm discovery accelerates practical applications. Quantinuum developed quantum algorithms on “the Hive” with Hiverge. The “Hive-ADAPT” method reduced quantum resource consumption by an order of magnitude compared to human-designed state-of-the-art versions.

Performance has improved due to Quantinuum and NVIDIA's cooperation. By adding NVIDIA GPU-based decoders to the Helios control engine, the group boosted logical fidelity by roughly 3%. This integration allows real-time quantum error correction decoding, which is crucial for scaling future systems. The Generative Quantum AI system ADAPT-GQE provided training data for complex compounds like imipramine 234 times faster.

The Future

A new collaborative endeavor is making the search for “quantum advantage” a public, high-stakes relay race in a world where technical discoveries often emerge privately. The Quantum Advantage Tracker, launched recently by luminaries in the field, is the core arena where quantum and classical computing approaches are in a “neck-and-neck” race few predicted this early. According to IBM Quantum researchers Jay Gambetta and Robert Davis, this open-source project is tracking progress and influencing quantum scientific validation.

Describe Quantum Volume.

IBM created Quantum Volume to measure quantum computers' true processing capacity. Unlike qubit counts, this metric considers hardware connections, error rates, and software compiler efficiency to assess performance. Finding the largest square circuit a machine can continuously operate yields an exponential number.

The Heavy Output Generation test ensures a device produces useful results rather than random noise. Despite application-specific workload constraints, it is a key industry benchmark for evaluating quantum computers' computational power. The paper stresses that hardware quality and durability, not scale, drive innovation.

Why Quantum Volume is the New Standard for Supercomputing Power in the Quantum Space Race Early in the quantum revolution, the industry sought qubit count. Tech businesses talked about 50, 72, or 100 qubit systems as if they were the only success factor. But as the field evolves in 2026, experts have found that outstanding performance doesn't always require many qubits.

If qubits are “noisy,” error-prone, or poorly linked, they cannot perform complex instructions. This revelation encouraged the industry to adopt Quantum Volume (QV), a more complete performance measure.

Beyond the Data: Orchestra Comparison

Professionals often utilize music to show why raw qubit figures are misleading. If qubit count matches the number of musicians on stage, a group can perform a difficult symphony without mistakes.

Ten world-class violinists are better than 100 non-players. Quantum volume considers qubit quality, gate fidelity, and coherence time.

Quantum Volume formula

Quantum Volume, a single benchmark figure representing actual capabilities, was introduced by IBM. It is measured using a “square” random quantum circuit with equal qubits (width) and operational steps (depth).

QV = 2k

Where k is the greatest number of qubits that may complete a square circuit, the simple yet effective formula is utilized.

To pass a test, the quantum computer must produce “heavy” outputs with a probability at least two-thirds higher than random guessing. A machine with a k-value of 10 maintaining a 10x10 circuit but failing at 11x11 has a Quantum Volume of 210, or 1,024.

How to Determine High Score?

High Quantum Volume is system-level, not hardware-level. Several aspects must work well together to boost the score:

Gate Fidelity: High error rates quickly impair calculations as the circuit deepens.

Connectivity is how easily qubits can "talk" across the semiconductor. Systems having “all-to-all” connectivity, such trapped-ion systems, perform better because they do not require additional “SWAP gates” that create errors.

Compiler efficiency depends on software stack. Intelligent compilers can enhance hardware performance by optimizing code for fewer steps.

Record and Competition Scene 2026

The quest for the largest Quantum Volume has become a high-stakes “space race” between technologies.

Trapped-ion systems lead this category. The company Quantinuum established a record of over 33.5 million quantum volumes in early 2026. Their H-Series systems, which reached 225 in September 2025, break records.

In the meantime, IBM advances superconducting technology. Their Heron r3 “Pittsburgh” system had 211 (2,048) QV in August 2025. Although easier to grow to huge qubit counts, superconducting devices often have more connectivity difficulties than trapped-ion devices.

Classical Simulation “Wall”

Quantum volume is approaching a fundamental limit despite its utilization. To validate a QV score, a traditional supercomputer must reproduce the circuit to verify the quantum machine's response.

After 50 qubits, ordinary computers cannot keep up with the simulation. The “proof” of scores for massive systems is nearly impossible. Layer Fidelity and CLOPS (Circuit Layer Operations Per Second) are being prioritized for the next generation of 1,000+ qubit CPUs.

The Future of Benchmarking

Although flawed, it uses fake random circuits instead of application workloads. Quantum Volume remains the most reliable indicator of progress in the NISQ era. Engineers are encouraged to build balanced machines rather than huge ones and loud systems are penalized.

The Quantum Decade: A ‘Vibe Shift’ in Fault-Tolerant Computing

The Quantum Computing Revolution

Academics estimated it would take decades to build a quantum computer that could perform exceedingly complex operations. Predicting chemical reactions for novel materials or understanding global communications' complicated encryption techniques are problems. According to Princeton University experimental quantum physicist Nathalie de Leon, the field is experiencing a “vibe shift” in the area. Perhaps in 10 years, high-performance quantum computers will exist.

Rapid improvement over the past two years gives new hope. Sources say teams from academia labs to large technology companies have reduced quantum system errors. Better hardware fabrication and operating practices for these vulnerable devices have achieved this. Computer scientist Dorit Aharonov of the Hebrew University of Jerusalem says we are in a “new era” where quantum computation is more likely and will arrive sooner than expected.

Overcoming Error

This shift is driven by fault-tolerant quantum computing. Quantum computers use qubits, which can be 0 or 1. The quantum spin of an electron, which can point in any direction, illustrates this. Entanglement, when multiple qubits become tightly correlated, increases information processing exponentially but makes the system vulnerable.

Quantum states drift and lose information, and qubit manipulation procedures like gates and measurements make mistakes, slowing development. Four separate teams recently revealed that these issues were resolved, a turning point. Google Quantum AI, Quantinuum, Harvard University with QuEra, and USTC implemented and improved quantum error correction.

One unit of “logical” information is spread across several “physical” qubits in this method. By monitoring physical qubits during a calculation, the machine may detect data degradation and fix it. The 1990s mathematical reasoning showed this was possible if mistakes stayed below a certain level; these four teams' recent performance proves this is possible.

Different Tech Directions

The race is reportedly taking place on multiple technology fronts. Google and USTC use superconducting material loops at slightly above absolute zero to protect electrons. In contrast, Quantinuum uses magnetic alignment of electrons within electromagnetic traps of ions. Meanwhile, QuEra uses light-based “optical tweezers” to manipulate neutral atoms.

These researchers want to reduce "overhead," the number of physical qubits needed to support one logical qubit. Scientists long believed this ratio must approach 1,000:1. Early estimates suggested billions of physical qubits were needed to factor enormous numbers, which was terrifying. However, new advancements are drastically reducing these numbers. Recent research by Google engineer Craig Gidney suggests that elaborate 3D geometric patterns in gate diagrams could cut factoring from 20 million to one million qubits.

The Efficiency Path

The current “name of the game” improves error correction. IBM claims a 100:1 ratio for encoding logical qubits with one-tenth the industry-standard overhead. QuEra is also investigating ways to use neutral atoms' flexibility to move and entangle, which could crack the 100:1 barrier. QuEra founder Mikhail Lukin thinks a gate integrity of 99.9%—known as the “three nines”—will enable this jump.

Experts like Nathalie de Leon research qubit metrology to remove noise. Her team increased qubit lives from 0.1 to 1.68 milliseconds by switching superconducting materials from aluminum to tantalum and employing insulating silicon instead of sapphire. De Leon believes lifetimes of 10 to 15 milliseconds are achievable, but removing one noise source usually reveals another.

New Post has been published on https://er10.kz/read/it-novosti/tehnologii-it-i-obrazovanie-obzor-sobytij-za-nedelju-16-01-26/

Технологии, IT и образование – обзор событий за неделю (16.01.26)

– Недропользователей в Казахстане обязали отчислять 1% на науку напрямую в бюджет

– Международные резервы Казахстана достигли почти $130 млрд

– Казахстанские дипломы будут признавать в 12 странах. Среди них: Австралия, Китай, Япония, Республика Корея, Россия, Новая Зеландия, Турция, Фиджи, Святой Престол, Монголия и Армения.

– Потребности экономики в электроэнергии полностью обеспечат в 2027 году – Минэнерго

– Производитель квантовых компьютеров Quantinuum готовится к IPO

– Выявлять теневые схемы в экономике правительству будет помогать ИИ. Для этого создадут цифровую платформу аналитики данных.

– Блокировать опасную рекламу в соцсетях предлагают в Мажилисе

– Госдепартамент США приостановил выдачу виз гражданам 75 стран

– В Казахстане зафиксирован рекордный рост числа иностранных студентов: на начало 2025/2026 учебного года в вузах страны учились 33,5 тыс. нерезидентов

– Казахстанцы массово выбирают вузы России, но меньше Кыргызстана и Беларуси

– ДУМК запустит «ИИ-имама». Приложение IMAM AI призвано обеспечить население быстрыми и надежными ответами на религиозные вопросы.

– Мобильные переводы в Казахстане: банки начнут передавать информацию налоговикам с апреля

– NASA впервые в истории провело экстренную медицинскую эвакуацию экипажа с МКС

– Актер Мэттью Макконахи зарегистрировал товарные знаки на свои мемные образы для защиты от ИИ

– Microsoft полнностью прекратила расширенную поддержку Windows Server 2008

– Платный браузер Opera One R3 получил «острова вкладок», звуковые темы и многое другое

– В Китае испытали ракету для суборбитального туризма – она слетала за линию Кармана и вернулась на Землю

– OpenAI выпустила онлайн-переводчик ChatGPT Translate

– ИИ-бот Qwen от Alibaba научился заказывать доставку еды и бронировать поездки

– TSMC может начать выпуск 3-нм ИИ-чипов OpenAI Titan уже в этом году – второе поколение перейдет на 1,6 нм

– Intel возвращает доверие инвесторов: акции компании приблизились к максимуму с 2023 года

– OpenAI договорилась о покупке ускорителей Cerebras на $10 млрд – чтобы снизить зависимость от Nvidia

– YouTube расширил родительский контроль – теперь можно заблокировать Shorts и не только

– Существующие функции Samsung Galaxy AI останутся бесплатными – платить придется за новые

– Вышел Mozilla Firefox 147 – улучшены аппаратное ускорение, управление вкладками и воспроизведение видео

– В ЕС назвали open source одним из компонентов цифрового суверенитета Европы

– Возрожденная ролевая песочница Hytale в духе Minecraft стала хитом – почти 3 миллиона одновременных игроков на старте раннего доступа

– DeepSeek научилась тренировать языковые ИИ-модели без оглядки на ограничения по скорости памяти

– ИИ-модели начали изучать как живые организмы – математика справляется все хуже

– ИИ-генератор видео Google Veo 3.1 научился создавать видео для TikTok и других вертикальных платформ

– Apple представила «новаторскую коллекцию мощных творческих приложений» Creator Studio

– Ошибка при обновлении ПО испортила 13 тыс. паспортов граждан Ирландии

– Метавселенная сожмется – Meta готовит массовые сокращения разработчиков Reality Labs

– Instagram исправил сбой с массовой рассылкой писем для сброса паролей, но причины утаил

– Пиковый онлайн Steam впервые в истории превысил 42 миллиона человек

– Китайские разработчики ИИ признают, что в ближайшие несколько лет им вряд ли удастся опередить США

– Google сделала рекламу в своих ИИ-инструментах для шоппинга более персонализированной

– В Windows 11 появится официальный способ удаления Copilot

– Миллионы наушников Sony, JBL, Xiaomi и других брендов оказались под угрозой взлома — в Google Fast Pair нашли уязвимость