Leading global experts indicate that the emergence of a universal quantum computer capable of swiftly solving real-world problems remains a long-term research challenge.
As artificial intelligence advances, a critical question arises: what direction should the next generation of computing and computational power take?
This question was central to the "Future Computing, Future Computing Power" forum at the 2026 World Artificial Intelligence Conference on July 18. Top international scholars emphasized that foundational breakthroughs in computing power represent long-cycle, cross-disciplinary scientific endeavors. Researchers must be willing to tackle problems that may not be solved within a single generation but require someone to begin the journey. Regarding quantum computing, participants noted that while China has achieved a series of milestone breakthroughs demonstrating quantum advantage, the development of a universal quantum computer that can quickly solve practical problems still requires sustained, long-term research. The academic community has a responsibility to guide the public in viewing quantum computing's development rationally and to guard against industry bubbles.
Chang Jin: Embracing the Challenge of Problems That May Span Generations
Chang Jin, President of the University of Science and Technology of China and an academician of the Chinese Academy of Sciences, drew from his experience in dark matter exploration. He proposed that the most valuable aspect of fundamental research is not necessarily finding immediate answers, but rather having the courage to confront challenges that may not be solved within one generation but must be initiated. He stated that breakthroughs in the foundational layers of computing power, like dark matter research, are long-term, interdisciplinary scientific pursuits. There are no standard answers deliverable in three years, nor can they be solved by a single discipline or institution. The physical limits of Moore's Law and constraints like energy, materials, cooling, and water resources are real challenges faced globally by academia and industry. The answers lie not in extending traditional paths but in exploring new directions like quantum computing, spatial computing, and novel physical principles. He reminded the audience that quantum information and green computing were niche fields twenty years ago. Their prominence today is due to young scholars who dedicated themselves to foundational work in the past two decades. The trajectory of computing over the next twenty years will similarly depend on the next, even younger generation.
The Future of Computing Power as a Utility and Its Four Key Challenges
In a dialogue with Guo Yike, Chief Vice President of the Hong Kong University of Science and Technology and a foreign academician of the Chinese Academy of Engineering, Gao Wen, Director of the Peng Cheng Laboratory and an academician of the Chinese Academy of Engineering, systematically outlined the logic, difficulties, and timeline for China's Computing Power Network. Gao Wen views the networking of computing power not as a new idea, but one that was previously impractical because supercomputing primarily served major scientific projects with limited users. Today is different, as computing power is increasingly used for AI training and inference—intelligent computing power. With the rise of the token economy, it has become relevant to the vast majority of people. He predicts that China's Computing Power Network will organize computing power through a network, and in the future, people will "use this computing power like electricity." Guo Yike described this scenario more concretely: in the future, individuals at home could use AI for health management, daily life organization, or even running a "one-person company," with computing power available on-demand like electricity.
This vision faces four major technical hurdles. The first is computing power supply. Gao Wen emphasized that true widespread application requires "integrated computing power"—primarily intelligent computing, but also incorporating supercomputing and general-purpose computing. Supply is inextricably linked to whether chip manufacturing and production can meet market demand, which currently falls short. The second is the network. He disagrees with the view that the existing internet is sufficient. Current networks only provide access guarantees; connecting computing power nodes over existing communication networks would be too costly, slow, and high-latency. A true computing power network requires connections at the data link layer, below the network layer, to reduce costs and latency. Technologies for cross-city, long-distance (hundreds to thousands of kilometers) bus-level data transmission and scheduling are currently weak or non-existent and need to be developed from scratch. The third is scheduling. Optimization must consider comprehensive costs including electricity and communication fees, not simply connecting resources haphazardly. The fourth is security, involving data security, privacy, and confidentiality. Beyond technology, there are market and application challenges: ensuring user satisfaction and forming a closed loop with the token economy.
On the cost side, Gao Wen listed four interrelated factors: first, chips, specifically the price of computing cards, which is the most watched factor; second, storage, often overlooked, but whose cost impact, if prices rise, could rival that of computing cards; third, communication costs, crucial after networking; and fourth, electricity costs, rarely considered in the past but essential for the long-term sustainability of a computing power center, requiring sufficiently low rates.
Regarding the timeline for practical use, Gao Wen divided the computing power network into three phases. Phase one is "aggregation," gathering national computing resources. This is about 70% complete, with 70% of national computing power now in a real-time updated database provided to the National Data Administration for monitoring the number of computing centers and their operational loads. Phase two involves collaborative computing for specific users (e.g., within companies or between partners, not the public) based on this aggregation, which will last several years, "likely around five more years." Phase three is opening to society after billing and management systems are fully in place, representing the true computing power network, which "may not be a matter of three to five years, but possibly a decade or two." He noted the current status: a small number of people are already using it.
Quantum Computing: Revolution or Bubble?
A roundtable titled "Exploring New Frontiers of Computing Power, Journeying to the Future of Computing" was moderated by Wang Jianping, Dean of the College of Computing at City University of Hong Kong. She opened by highlighting the core dilemma: on one hand, quantum computing hardware is advancing rapidly, yet a universal quantum computer seems distant; on the other, the market is eager to capitalize on this wave, but technical paths haven't converged, and commercial pathways are unclear. Is today's discussion about quantum computing a revolution in progress or a long journey full of unknowns? Five panelists offered their perspectives.
Lu Chaoyang, Professor at the University of Science and Technology of China and Executive Dean of the Shanghai Institute for Advanced Studies, first clarified the significance of "Jiuzhang." This was the first time a quantum machine was built using a completely new principle, enabling it to surpass classical computers on specific tasks. It cannot comprehensively defeat classical computers, but for the first time, it "somewhat like AlphaGo" demonstrated the computational power of quantum computing and confirmed the conjecture of theoretical scientists like Feynman—whether quantum mechanics could provide an acceleration. He then corrected a common misconception: hearing "universal" leads people to think it can compute anything and replace classical computers, which is not the case. Even if a future universal machine is built, it can only accelerate certain specialized problems. The real difference between a universal and a specialized machine is not literal but lies in the near absence of noise or its extreme suppression in a universal machine, whereas specialized machines lack error correction.
Discussing technical routes, Lu Chaoyang believes it's not yet time for convergence. Paths like photonic, superconducting, ion trap, and neutral atom are all viable with sufficiently talented people. The key is not which route is chosen or whether enough funding can be raised, but whether a sufficiently excellent and collaborative team can be assembled. In the next three to five years, preliminary, fault-tolerant universal quantum computing prototypes are likely—about 100 to 1000 logical qubits with error rates as low as 10⁻⁶ or 10⁻⁹. While "small but complete" and not yet capable of breaking encryption, they could find initial applications in scientific discovery problems. He emphasized that the success of any technical route depends on two things: the principle being viable, and gathering people dedicated to solving problems pragmatically, not fabricating stories for fundraising or falsifying resumes.
Zhao Wei, Provost of Shenzhen Institute of Technology and an academician of the International Eurasian Academy of Sciences, first clarified concepts. He stated that quantum science is one of humanity's greatest discoveries of the past century, and young people should seriously study its basic concepts—"viewing the world through quantum science." Quantum science enables quantum technology, divided into quantum information, quantum communication, and quantum computing.
He emphasized that if a quantum computer is compared to a car, we have now built the "engine"—even if just a single-cylinder one, with the potential for two, three, sixteen, or even hundreds of cylinders in the future. However, "having an engine doesn't mean you have a car"; a complete system and architecture are still needed on top. He elaborated that Turing precisely defined digital computation, but it was von Neumann who turned digital computation into the digital computer. Today's universal or semi-universal quantum computers lack a pioneering figure akin to von Neumann in the digital computer era, or a publicly recognized, viable architecture and supporting system. "Everything we do today is preparing for the emergence of such a person or team."
Ying Yong, President of QuantumCTek, divided the issue into two timelines. The first is technological evolution: Phase one is achieving quantum advantage, corresponding to Google's 2019 achievement, Lu Chaoyang's "Jiuzhang" in 2020, and Zhu Xiaobo's "Zuchongzhi" series in 2021, which validated it on specific problems. Phase two, in the next three to five years, involves specialized quantum computers, aiming for quantum advantage on problems with practical value, albeit with limited computing power and convenience. Phase three is universal programmability. Only after solving error correction and running algorithms like code-breaking will it bring revolutionary changes to production and life, requiring another decade or more.
The second timeline is the commercialization and industrialization of applications. Ying Yong stated frankly that industrialization requires a sizable market and strong supply and demand, which quantum computing clearly hasn't reached. Is there commercial value before industrialization? He believes it's certain. He revealed that quantum computing accounted for about 40% of his company's sales revenue last year, including whole machine sales, component sales, and quantum computing cloud services. However, current users purchase machines and cloud services not for code-breaking, weather forecasting, or material/drug design, but more for algorithm research, hardware research, talent training, and science education. Demand in these areas "is visibly growing." His conclusion: the door to quantum computing commercialization is open, but the path to scaled industrialization is long.
Qiu Dagen, Legislative Council Member (Functional Constituency - Technology and Innovation) of the Hong Kong Special Administrative Region, approached from Hong Kong's position. He said Hong Kong is preparing its first five-year plan, having collected much market feedback, with quantum being a major focus. Regarding talent, Hong Kong has five world top-100 universities. As a national talent hub, the SAR government has invested considerable resources in "from 0 to 1" source innovation in recent years, attracting several internationally renowned professors, with quantum being a key focus area.
Financial security was his main point. As an international financial center, Hong Kong handles resident deposits, funds under management, and stock market capitalization in the trillions of dollars, with most financial systems still using traditional RSA/ECC encryption. If quantum computing breaks through, whether the financial system can maintain security and resilience directly impacts its status as an international financial center.
He noted that the Hong Kong Monetary Authority issued guidance last year to financial institutions on post-quantum cryptography preparation, and Hong Kong has a "Fintech 2030" strategy. In his view, Hong Kong's positioning is to "provide standards": regardless of how various countries' technologies develop, systems ultimately need to interconnect and communicate. This requires hybrid models incorporating both domestic and international algorithms, with standards needing international consistency and connectivity—a direction Hong Kong can participate in. Qiu Dagen stated that the trend from the internet to blockchain, to AI, and quantum will only move forward, with speed beyond imagination, and everyone must enhance their understanding of quantum.
Enkhbayar Nambar, the third President of Mongolia and Chairman of the New Kharkhorin City Council, stated that Mongolia is a "mobility-based" country. About 30% of the population are nomads, close to nature and adaptable to external changes. As quantum computing enters life, bringing rapid changes, he believes Mongolians will keep pace. He expressed hope to invite scholars from China and other countries to Mongolia to test and validate innovative results, for which special preferential policies have been prepared.
When moderator Wang Jianping asked each panelist for a one-sentence answer, the five focused on different aspects. Zhao Wei said quantum technology development won't be smooth sailing; the entire industry must dare to take responsibility for a bright future. Lu Chaoyang advised young people to "learn to appreciate and understand the beauty of quantum mechanics," while urging financial institutions to "quickly migrate to quantum cryptography" because the quantum race will progress faster than imagined. Ying Yong returned to the most practical point—the current focus should be on significantly improving the hardware level of quantum computers.
Pan Jianwei: Academia's Role in Rational Guidance and Bubble Prevention
Pan Jianwei, an academician of the Chinese Academy of Sciences and Executive Vice President of the University of Science and Technology of China, set the tone for the "Towards Quantum" segment. He stated that emerging quantum information technology is flourishing and will bring breakthroughs in privacy protection, sustainable computing power growth, public safety, life and health, potentially driving another great leap in human civilization. Quantum computing originated from Feynman's idea of "using controllable qubits to simulate complex quantum systems." It can not only quickly break certain classical codes but also help humanity deeply understand nature, greatly advancing chemistry, materials science, and life sciences.
He also soberly noted: "Despite achieving a series of milestones in quantum advantage, a universal quantum computer capable of quickly solving practical problems still requires long-term research. Therefore, sustained investment from both government and the private sector is essential. On the other hand, the academic community has a responsibility to guide the public in viewing quantum computing development rationally and to guard against industry bubbles." He also quoted Heisenberg, emphasizing that international academic exchange remains an indispensable driver of scientific progress.
Franco Nori: Quantum Computing as a Marathon, Not a Sprint
Franco Nori, Chief Scientist at Japan's RIKEN and a professor at the University of Michigan, repeatedly emphasized an analogy: quantum computing is more like a marathon, even a 100-kilometer ultramarathon, not a 100-meter sprint. The prospects are very bright, potentially revolutionizing materials science, chemistry, cryptography, machine learning, and medicine, but the most important and profitable applications likely won't arrive quickly. Timelines for the latter vary, with estimates of ten, twenty, or thirty years. "The problem is precisely that many investors treat it as a 100-meter sprint, which breeds bubbles."
He said the previous internet bubble lost about $5 trillion, while the current AI bubble is much larger, "estimated to be 17 times that of the internet bubble." In comparison, the quantum bubble is much smaller but still requires vigilance. Nori does not deny the technology itself—participation by giants like Google, Microsoft, and IBM has yielded results; the technology is reliable and scientifically sound. "The problem lies on the investment side," where there is excessive marketing. He condensed his judgment into a simple question: if you remove all marketing videos, company press releases, and promotional materials and ask, "What is actually being sold today? What useful, practical, profitable real-world problems are being solved?"—currently, there are only some preliminary yet exciting results, insufficient to support such investment bubbles. He reminded that history shows repeated "hype cycles": innovation explosion, peak expectations, subsequent disappointment, and finally, discovery of what truly works.
Zhu Xiaobo: The Quantum Computer Endeavor as a "Mars Landing Plan"
Zhu Xiaobo, Professor at the University of Science and Technology of China and Chief Designer of the "Zuchongzhi" quantum computing prototype, first explained the principle: quantum computing is based on qubits; as the number of qubits increases, the state space grows exponentially. A quantum computer with over 70 qubits holds information that all the world's hard drives couldn't store, yet its input/output is very inefficient. In terms of applications, code-breaking has the clearest path, knowing what kind of machine and performance level is needed to crack mainstream asymmetric encryption. Quantum chemistry and quantum simulation were Feynman's original vision. Regarding AI, he is personally optimistic but admits, "So far, there is no clear plan for how to apply it precisely, accurately, and step-by-step to AI," only numerous exploratory results indicating "it's very likely to happen."
Why will a universal fault-tolerant quantum computer take another 10 to 15 years? Zhu Xiaobo laid out the reasons. We are manipulating individual quanta; the energy required to flip a superconducting qubit is "about one hundred quintillionth of the energy released by a butterfly flapping its wings." Any slight disturbance drastically increases the error rate. At current technological levels, achieving an error rate of one per thousand is near the technical limit. Based on the mainstream superconducting route, the world's best level is a flip error rate on the order of 10⁻³ with hundreds of qubits. In contrast, classical computers typically have error rates of 10⁻¹⁵ to 10⁻¹⁸. He warned, "Don't believe any hype about 'achieving thousands or tens of thousands of qubits.' Simply increasing the number of qubits is meaningless if the precision and error rate aren't good enough."
His proposed path: achieve thousands/tens of thousands of qubits with single-qubit error rates reduced to one per thousand in about 5 years. Then spend 5 to 10 years constructing a universal fault-tolerant quantum computer using hundreds to thousands of logical qubits. Among the three stages, stage one (surpassing classical computing on specific problems) currently includes only Google's "Sycamore" and China's "Zuchongzhi-3" in superconducting, and China's "Jiuzhang" and a Canadian company in photonics, with ion traps recently achieving it as well. Stage two involves solving problems with application value (e.g., quantum chemistry of large atoms/molecules),有望在未来几年实现并可小范围应用.
If the quantum computer endeavor is compared to a "Mars landing plan"—we are currently in the ground experiment stage, aiming to conquer Mount Everest. He said that overhyping quantum computing's applicability in finance and biology now "is all boasting," which would only deal a devastating blow to the entire field. "Hype is all harm and no benefit to our field."
In a subsequent dialogue with moderator Lin Jing, Professor at the City University of Hong Kong's College of Computing, Zhu Xiaobo clarified the terminology of "quantum advantage." The term originally from the US was "quantum supremacy," but Chinese teams, from their earliest publications on Jiuzhang and Zuchongzhi-3 in 2020 and 2021, used "quantum advantage" rather than "supremacy."
He said quantum advantage only proves in principle that quantum computers can surpass classical computers on specific problems, which is scientifically significant but far from practical application. Finding useful problems for the second stage is much harder than imagined, primarily due to the high error rate—algorithm experts have proposed many ingenious algorithms, but on average, only 10 to 20 single-gate operations can be performed per qubit before "guarantees are lost and it's all errors."
He judges that quantum simulation is currently more promising,有望在未来几年在难解的大分子体系上取得突破. Code-breaking必须依靠通用容错量子计算机, as the number of gates is on the order of 10⁸, making it基本不可能 unless error rates are reduced to 10⁻⁸ to 10⁻¹⁰. As for AI, a neutral assessment is that it至少需要 "early fault-tolerant quantum computers"—systems with error rates reduced to one per million and hundreds of qubits—to possibly succeed.
Comments