2026 is crucial for QPU to replace GPU.
The arrival of the QPU era may be faster than anyone expected.
Pat Gelsinger, the former CEO of Intel, said in a recent interview that quantum computing will become popular within two years and accelerate the bursting of the AI bubble, and will completely replace GPUs before 2030. In his view, quantum computing, classical computing, and AI computing will together form the "holy trinity" of the future computing world.
On December 9th, QuantWare, a quantum hardware startup, officially released a new quantum processor (QPU) expansion architecture, VIO - 40K, and plans to achieve mass production of quantum chips in 2026. At its headquarters in Delft, the Netherlands, the company is actively building Kilofab - the world's largest and first wafer fab dedicated to quantum chip production. It is reported that this will increase its production capacity by 20 times compared to the current level.
At the beginning of 2025, Jensen Huang joked that "quantum technology is at least 20 years away from practical application." However, by the end of the year, the roadmaps of various quantum startups and giants were becoming more and more aggressive, and the predictions of the media and institutions also began to turn optimistic.
Undoubtedly, 2026 will be a crucial year for quantum computing to move towards practical application. And now is a good time to sort out the technology and review the market.
Let's first take a look at what happened in the quantum industry in 2025.
01 The Quantum Explosion in 2025
In 2025, the main players on the quantum stage can be summarized as: three tech giants, the "Quantum Four", and NVIDIA.
Google: After releasing the Willow superconducting quantum processor with 105 physical qubits at the end of 2024, Google officially announced the achievement of "verifiable quantum advantage" in October 2025. By running the "Quantum Echo" algorithm, the Willow chip was 13,000 times faster than classical supercomputers in processing out - of - order time correlator tasks. Experimental data further showed that as the scale of physical qubits increased, the logical error rate of the system decreased exponentially, verifying the effectiveness of the error - correction theory at the physical level.
In terms of the ecosystem, Google deepened its cooperation with NVIDIA, using the CUDA - Q platform for large - scale physical simulations to solve the noise design problems of next - generation processors. In addition, Google reached a cooperation with the UK National Quantum Computing Centre (NQCC), opening cloud access to the Willow chip for UK research institutions to support algorithm testing in fields such as materials science.
IBM: In 2025, IBM continued to implement its hardware iteration roadmap, focusing on improving processor performance and reducing the cost of the control system. In terms of hardware, IBM released and delivered a 120 - qubit processor codenamed "Nighthawk". This chip uses a new - generation tunable coupler, with a computing performance approximately 20% higher than its predecessor, "Heron". At the same time, IBM introduced an experimental chip called "Loon" to verify the stability of large - scale fault - tolerant components.
In terms of the engineering control architecture, IBM announced the results of its cooperation with AMD in October of the same year. The two sides used AMD's existing commercial FPGA chips to achieve real - time error - correction control of qubits. Tests showed that the processing speed of this solution met the requirements for real - time error correction, and the project progress was completed one year ahead of schedule. This progress demonstrated that general - purpose commercial chips can replace customized hardware for quantum control, which helps reduce the engineering cost of building fault - tolerant quantum computers and supports IBM's plan for fault - tolerant models in 2029.
Microsoft: In February 2025, Microsoft released its first quantum chip, "Majorana 1", based on topological superconductor materials. This product marked that Microsoft's long - term investment in the topological quantum computing route had entered the hardware prototype stage from theoretical research. Microsoft stated that the chip uses the characteristics of new materials to achieve immunity to environmental noise at the physical level. Although the current qubit scale is small, the release of this prototype verified the feasibility of the topological protection mechanism at the hardware level, providing an experimental basis for subsequent expansion.
The "Quantum Four" (IonQ, Rigetti, D - Wave, QCI): In 2025, listed quantum computing companies represented by IonQ, Rigetti, D - Wave, and Quantum Computing (QCI) underwent market integration and business adjustments. IonQ completed the acquisition of UK startup Oxford Ionics in June, involving an amount of approximately $1.1 billion, aiming to integrate ion - trap technology patents and expand the engineering team. Rigetti Computing continued to advance the modular deployment of superconducting systems and optimized the interconnection performance of the Ankaa - 3 processor. D - Wave focused on demonstrating the hybrid solving ability of quantum annealing technology in logistics scheduling and supply - chain optimization. In addition, Quantum Computing continued to explore in the field of photonics, aiming to lower the environmental threshold for system operation.
NVIDIA: In 2025, NVIDIA underwent a strategic correction in the quantum field. From doubting quantum computing at the beginning of the year, Jensen Huang publicly apologized at the GTC Summit in March and established the strategic position of quantum computing. Subsequently, NVIDIA quickly initiated capital operations.
In September, NVentures, a subsidiary of NVIDIA, continuously made significant investments in three technology routes within a week: participating in the nearly $600 million financing of Quantinuum (ion - trap), supporting QuEra (neutral atoms), and following the $1 billion Series E financing of PsiQuantum (photonic quantum). This combination of moves aimed to cover the mainstream hardware modalities through capital rather than self - developing QPUs.
At the GTC Conference in Washington at the end of the year, NVIDIA launched NVQLink, enabling direct communication between the QPU of quantum computers and GPUs. Jensen Huang pointed out that the QPU of quantum computing has received unprecedented support, and currently, 17 quantum computing companies and 8 US Department of Energy (DOE) national laboratories have joined the NVIDIA ecosystem.
02 A Series of Highlights in 2026
Based on the technology roadmaps of major manufacturers and the predictions of industry analysis institutions, 2026 is regarded as a crucial node for quantum computing to move from engineering verification to utility verification. The industry's focus will shift from simply expanding the scale of physical qubits to verifying the quality of logical qubits and the actual deployment of hybrid computing architectures.
In the superconducting route, IBM has set 2026 as a crucial year to demonstrate "quantum advantage" in its plan. The company plans to use a "quantum - centric supercomputing" architecture to try to prove its cost or accuracy advantage over classical computers in specific scientific tasks. Its processor is expected to support deeper - level quantum gate operations (targeting about 7,500 gates) to run more complex algorithms. Google, on the other hand, faces the engineering challenge of transitioning from verifying error - correction principles to building long - lived logical qubits, focusing on further improving the coherence time and gate fidelity of physical qubits.
In the emerging routes, the neutral - atom manufacturer QuEra Computing has set the goal of releasing a system with 100 logical qubits; the photonic - quantum manufacturer PsiQuantum is accelerating the assembly of its ultra - large - scale systems in Chicago and Brisbane and is expected to enter a crucial system integration stage in 2026; and QuantWare has launched the plan for the Kilofab wafer fab, attempting to promote the industrial mass production of quantum chips. All of the above plans have high engineering uncertainties, and 2026 will be a test period to verify whether these aggressive goals can be achieved.
With the promotion of middleware platforms such as NVIDIA's CUDA - Q, "Quantum - classical hybrid computing" may become the standard deployment mode in data centers in 2026. QPUs will be more integrated as accelerators into high - performance computing clusters to handle specific simulation or optimization tasks. The popularization of this architecture will promote the combination of quantum computing and AI workflows, especially in the fields of large - model training optimization and complex molecular simulation. The industry expects to see more application tests based on hybrid computing power.
In the field of cloud services, IBM, AWS, and Microsoft currently provide quantum access services. In 2026, cloud providers may further integrate "quantum + classical" hybrid resources. The service model will evolve from providing single experimental hardware access to providing a hybrid computing environment integrated with high - performance computing resources. The threshold for enterprise users to call quantum computing power through the cloud platform to solve practical problems is expected to be lowered, but whether it can generate a positive return on investment in large - scale commercial scenarios remains to be seen.
At the security level, the US government is forcing the anti - quantum upgrade of digital infrastructure. According to the strategic deployment of the National Security Memorandum (NSM - 10), US federal agencies have entered the actual implementation stage of post - quantum cryptography (PQC) migration. The White House requires agencies to accelerate the phasing out of traditional encryption algorithms such as RSA and has set 2025 - 2030 as the critical window period for core systems to complete PQC upgrades. This initiative aims to establish the US's leading position in formulating the next - generation encryption standard and drive the reconstruction of defense systems in key fields such as finance and national defense.
This also reflects the fact that the quantum computing competition in 2026 will not only take place at the enterprise level but also between major countries around the world.
03 Countries Are Building Quantum Industrial Systems
Looking forward to 2026, as major global economies enter a new policy cycle, the development logic of quantum computing is undergoing profound changes. The focus of countries is gradually shifting from the early competition in single scientific research indicators to building a resilient industrial system and mastering the core links of the industrial chain.
The US strategic focus in 2026 has shifted to strengthening its internal industrial foundation. In response to the labor - force gap risk pointed out by the non - profit organization "QED - C", the National Institute of Standards and Technology (NIST) is expected to accelerate the cultivation of quantum industry clusters in places such as Colorado. Different from mainly funding basic scientific research in the past, the R & D budget orientation for the 2027 fiscal year shows that US policies are more focused on cultivating skilled technical workers and engineers and exploring new support models, including equity - for - capital, aiming to build a complete local industrial closed - loop from R & D to manufacturing to address the talent shortage in the fields of cryogenic electronics and microwave engineering.
2026 is the starting year of China's 15th Five - Year Plan, and quantum technology is positioned as a "new economic growth point" in the relevant planning suggestions, which implies that China's quantum industry is accelerating the transition from laboratory verification to industrial cultivation. At the infrastructure level, China's industrial community is exploring the integration of quantum computing into the national - level computing power network, relying on the "East - West Computing" project. China Telecom Quantum and Origin Quantum and other enterprises are expected to further pilot the "integration of four types of computing", namely quantum + supercomputing + intelligent computing + general computing, trying to use the real needs of public scenarios such as weather forecasting and power grid scheduling to drive technological iteration.
In terms of industrial chain construction, Chinese enterprises are committed to building a full - stack autonomous ecosystem. Manufacturers represented by Liangi Technology are promoting the R & D and verification of key components such as dilution refrigerators and special cables. At the same time, the photonic - quantum route is also being accelerated. Turing Quantum and other complete - machine manufacturers are exploring differentiated applications in the fields of AI and biomedicine, while Tengjing Technology provides solid underlying hardware support for this route through the R & D of precision optical components.
Europe and the UK are also accelerating the implementation of the "technological sovereignty" strategy. The UK has promised to continuously invest funds to support the quantum industry, relying on domestic enterprises such as ORCA Computing to build an internationally competitive industrial high - ground. The EU tends to support domestic companies in building quantum chip foundry facilities, trying to gain more say in the chip manufacturing link and build an industrial "third pole" with independent innovation capabilities.
This article is from the WeChat official account "Semiconductor Industry Insights" (ID: ICViews), author: Semiconductor Industry Insights. Republished by 36Kr with permission.