人工智能（AI）与量子计算之间的关系在竞争与合作中不断演进。早在2013年，谷歌和美国国家航空航天局（NASA）就建立合作实验室，尝试利用D-Wave的量子计算机优化机器学习算法，但当时研究显示其性能并未超越传统的图形处理器（GPU）。如今，在生物、化学和材料科学等领域，AI公司如Isomorphic Labs和CuspAI正尝试通过模式识别来预测分子特性，而量子计算机则致力于提供高精度的物理仿真。尽管中国研究人员在3月展示了量子系统在天气预报中超越传统方法的潜力，但IBM专家指出，两者并非零和博弈，量子计算机生成的高精度仿真数据能为AI模型提供更优质的训练集。
在技术构建层面，AI与量子计算正形成强有力的双向赋能。面对量子比特（qubits）极易受干扰且易出错的难题，初创公司Infleqtion推出了由AI驱动的“解码器”，帮助量子计算机进行高效的错误纠正。与此同时，量子计算机也为AI模型训练提供了新型硬件加速方案。澳大利亚公司SQC推出的量子蓄水池芯片（quantum reservoir chips）被电信公司Telstra采用后，成功将AI模型的训练时间缩短了90%。IBM也正探索将量子计算机应用于现代AI核心计算中，以替代部分耗电巨大的GPU集群。
在资源与能源效率方面，量子计算展现出独特的优势。虽然高级量子计算机需要在接近绝对零度（如-269°C）的液氦环境中运行，但其整体能效远优于需要吉瓦（GW）级电力支持的超级AI数据中心。例如，IBM的量子系统能耗仅停留在兆瓦（MW）级别，而SQC的硬件可直接安装于标准服务器机柜中，功耗不足同等GPU机柜的十分之一。随着美国政府于5月向IBM、D-Wave等量子实验室投资20亿美元，AI与量子计算的深度融合正加速推向实用化阶段。
The intersection of artificial intelligence and quantum computing reflects a subtle balance between competition and synergy. In 2013, Google and NASA established a joint laboratory using D-Wave quantum hardware to optimize machine learning, though initial benchmarks revealed no advantage over conventional graphics processing units (GPUs). Today, AI enterprises like Isomorphic Labs and CuspAI employ pattern-recognition models for molecular and chemical prediction, directly competing with quantum platforms built for first-principles physical simulation. However, industry experts at IBM emphasize that this is not a zero-sum dynamic; quantum systems can generate highly accurate baseline simulation data to train more capable AI models.
Technologically, AI and quantum architectures provide mutual acceleration. To mitigate the inherent instability of physical qubits, startups like Infleqtion deploy AI-driven decoding algorithms that assist real-time error-correction logic alongside quantum hardware. Conversely, quantum computers are providing novel acceleration mechanisms for AI models. Australian firm SQC developed quantum reservoir processor chips that allowed telecommunications operator Telstra to reduce AI model training durations by 90%. Furthermore, IBM is actively researching methods to integrate quantum processors into core machine-learning pipelines, providing alternatives to traditional GPU-intensive infrastructure.
From an energy and resource perspective, quantum platforms present compelling efficiency advantages over massive AI data centers. Although systems like IBM’s System Two require liquid-helium cooling to near absolute zero (-269°C), their power demand operates in the megawatt range rather than the gigawatt scale demanded by modern AI superclusters. For instance, SQC’s rack-mountable quantum hardware consumes less than 10% of the energy of an equivalent GPU server rack. Supported by a $2bn U.S. government equity investment in May targeting firms like IBM and D-Wave, the integration of AI models and quantum computing is transitioning into practical deployment.
Source: AI and quantum computers will be frenemies
Subtitle: The two technologies look more complementary than rivalrous
Dateline: 7月 30, 2026 05:36 上午