AI-driven computing anxiety is pushing industries and capital markets to explore fundamental shifts in computing paradigms. "We believe that traditional computing expansion paths face dual constraints of cost and energy consumption, while quantum computing offers a new possibility," said Ju Jiangwei, Vice President of Beijing Boson Quantum Technology Co., Ltd. (hereinafter "Boson Quantum"), in a recent interview.
Boson Quantum's VP stated, "Quantum computing is an essential question for AI, and vice versa." However, he clarified that this does not imply a replacement or challenge to classical computing, but rather integration and symbiosis. "Quantum AI and quantum-classical hybrid computing are essentially the same thing."
In November 2025, Boson Quantum established China's first dedicated quantum computer manufacturing factory in Shenzhen, delivering products as engineered systems to commercial clients. Its solutions are being explored across AI, communications, finance, medicine, energy, and new materials discovery. In July, the company filed for an initial public offering (IPO) guidance filing. Quantum computing is accelerating its move from labs to real-world scenarios. However, amid rising industry excitement, Ju emphasized that the commercial deployment of quantum computers remains in its early stages. The search for a "killer application," the development of the industry's "middle layer," and a scarcity of cross-disciplinary talent represent more tangible ecosystem challenges. "When the industry chain is immature, customers must step back a half-step, and we must step forward a half-step," Ju explained.
Where to begin?
Ju noted that the quantum industry's development and change are very evident regarding market acceptance and maturity. From 2020 to 2023, market expansion required first explaining what quantum computing is, but since 2024, especially last year, a significant shift occurred where clients are now well-informed and focused on how to use quantum computers. The rising industry heat and market recognition are driven by policy, but the core reason is a positive feedback loop from industry development and demand-side pull. The deep integration of AI and large models across industries is fueling a growing demand for computing power, highlighting the limitations of classical computing. Consequently, clients are experimenting with new computing technologies, including quantum computing. It has already shown the ability to solve practical problems in specific scenarios and is expanding to more applications.
Is quantum computing a choice or a necessity for AI?
Ju explained that the primary advantage of quantum computers stems from the underlying properties of quantum physics, enabling parallel accelerated computing to solve large-scale, complex problems with benefits in time, efficiency, or energy consumption. This is an inherent advantage of its physical system, not something achievable by simply "stacking" classical computers. However, at the hardware level, quantum computing is not yet mature enough for large-scale applications. It is more about combining the strengths of classical and quantum computing to achieve a "1+1>2" effect within specific applications. This is a gradual fusion process of new and classical computing power, not a simple replacement, even in the long term.
Will quantum computing disrupt the current AI ecosystem?
Boson Quantum determined from its inception that regardless of future applications, AI would be a primary battleground for quantum computing. Therefore, the company focused its quantum computer's main direction on AI from the start. Ju stated that the current AI "tree" grown on the architecture of backpropagation algorithms, large models, and GPUs is not something quantum computers seek to challenge but to empower. The 2024 Nobel Prize in Physics awarded to Hinton and Hopfield, for their contributions to Boltzmann neural networks, aligns well with Boson Quantum's dedicated quantum computers. The core path for applying dedicated quantum computers to AI for Science is "Quantum for AI for Science," where the "AI" refers more to Boltzmann neural networks than current generative AI represented by Transformers. On one hand, Boltzmann neural networks and deep learning are parallel architectures; on the other hand, embedding Boltzmann sampling into deep learning can enhance it by integrating with diffusion models and Transformer architectures. Boson Quantum calls these two directions "physical-native AI" and "physical-enhanced AI," which is the technical path they have identified for quantum computers to enter the AI field.
Has the quantum computing industry's "temperature" risen?
Ju noted that research-oriented users still dominate, even among industry users, it is often those focused on research within their sectors. This research is not purely academic but aimed at exploring industrial applications. Finding scenarios for quantum computers is a two-way effort between the company and the industry.
Why is finding scenarios for quantum computers so difficult?
Ju explained that any new technology undergoes this process, and even large models are still finding their scenarios today. Since quantum computers are primarily positioned for B2B and government clients, solving problems for large state-owned enterprises, research institutes, and universities, scenario discovery is inherently challenging. Furthermore, there are industry chain development issues. Building a quantum computer is only the first step. Like gaming, you need hardware, an operating system, apps, and a network connection—this is a complete industry chain that no single company can build. It involves multiple products at different levels to meet user needs. While the technological gap between China and the West in quantum computing is not huge, the latter is more advanced in industrialization. For example, Western quantum computing companies have begun to form a differentiated industry chain, with unicorns emerging in areas like quantum computer manufacturing, middleware, and application software, alongside several listed companies. In China, the quantum computing industry chain has not yet formed clear differentiation, and the capitalization process lags behind the US. However, over the past year, more young teams have emerged in the primary market, focusing on hardware, software, and algorithms, gaining recognition and support. It is evident that China's quantum technology sector is also increasing investment and progress.
What are the key milestones for quantum computing industrialization?
Ju outlined the first major milestone as finding a practical scenario that truly leverages the value of quantum computers, even if its computing power matches classical computers but with lower power consumption, making it worth paying for. The second stage is quantum-classical hybrid computing. Quantum AI and quantum-classical hybrid computing are essentially the same thing, one being a combination at the algorithm level and the other at the infrastructure level. The future will involve multiple computing powers integrating to solve complex problems, with quantum computers being just one of them, not something to be mythologized.
How does Boson Quantum choose what to do and what not to do?
Boson Quantum has a clear positioning as a quantum computer hardware company centered on "practical quantum computing." It focuses on hardware, continuously improving stability and qubit scale, and on developing the quantum neural network development tool Kaiwu-PyTorch-Plugin (KPP), allowing developers without quantum physics backgrounds to use quantum computing power for applications, lowering the ecosystem barrier. Additionally, Boson Quantum advances scenario exploration through campus quantum competitions, research funds, and corporate partnerships. When the industry chain is immature, customers must step back a half-step, and the company steps forward a half-step, conducting scenario exploration and incubation of midstream enterprises.
How does Boson Quantum view different technology routes?
Ju noted that quantum computer technology routes include superconducting, ion trap, photonic, and neutral atom. Currently, no single route has converged; it is hard to say which is better, and they are progressing in parallel. It is possible that multiple routes will coexist in the future, each with its own pros and cons for different scenarios. From a business perspective, the goal is to produce a quantum computer that solves problems. If it cannot solve a problem, no route is effective. Academic research's main battlefield is the lab, while companies must face real industry needs. Before technology fully converges, companies must maximize the advantages of different routes and minimize disadvantages. The main advantage of photonic quantum computing is its ability to prepare quantum states at room temperature, leveraging a mature optical industry chain. The talent pool, industry chain maturity, and room-temperature operation give photonic quantum computing an edge in engineering deployment for dedicated quantum computers, with manageable user maintenance, operational, and purchase costs. Compared to general-purpose quantum computers, dedicated quantum computers have a narrower application scope, focusing on specific problems, but their qubit scale and stability allow users to start using them in these areas. General-purpose quantum computers need more time to mature from research to commercial use. This does not mean dedicated quantum computers are better overall, but they offer a more reasonable technology for users at a specific stage.
What is the biggest non-technical risk for Boson Quantum's future?
Ju identified the recruitment and retention of top talent. The industry now lacks interdisciplinary and engineering talent. The phase of finding quantum physics experts to build a quantum computer is over; the challenge now is how to use it, requiring cross-disciplinary collaboration. The company looks for candidates with backgrounds in biology, finance, materials, and quantum physics. Many downstream clients are also hiring and training cross-disciplinary talent, who are the ones capable of developing the quantum algorithms needed for specific applications. As the industry develops, talent bottlenecks will slow the pace of algorithm application deployment. Solving this requires the ecosystem. It is not just a quantitative gap but a structural issue involving the education system, training cycles, and global competition.
Comments