By Xinhua writer Xia Xiao
DUBAI, Feb. 27 (Xinhua) -- High-performance computing (HPC) has evolved from a specialized tool into a core pillar of national scientific capability in the era of artificial intelligence and big data, Turing Award laureate Jack Dongarra said.
"I see high-performance computing as part of a nation's scientific 'instrumentation,' alongside telescopes, light sources and accelerators," Dongarra said in a recent interview with Xinhua. "It's how we turn theory, data and experiments into predictive capability."
As computing becomes strategic infrastructure, building national HPC capacity involves more than acquiring powerful machines, he said.
"The hardest part is building a capability that delivers time-to-science," Dongarra said, referring to the ability to translate computing power into timely scientific results. That requires clear mission focus -- whether for large-scale simulation, AI training, data analytics or digital twins -- as well as reliable power, cooling and operational capacity at scale.
Dongarra said the boundary between HPC and AI is rapidly blurring.
"AI and HPC aren't two separate worlds anymore," he said, noting that AI techniques are increasingly embedded in scientific applications, while HPC systems underpin large-scale AI training and inference.
He said the most significant integration will come in AI-for-science, where physics-based models, data and machine learning reinforce one another. Success, he added, depends on linking models to real-world data and validating results at scale.
As countries treat computing as strategic infrastructure, geopolitical factors are reshaping technology supply chains.
"A country needs enough autonomy to ensure resilience, especially for critical workloads, but it also needs openness to stay at the scientific frontier," Dongarra said. "If you close everything, you slow down. If you open everything without policy, you invite risk."
Asked about China's recent progress in HPC, Dongarra gave a positive assessment.
He said China has treated HPC as a strategic national capability "systematically and at scale," supported by sustained long-term investment, close alignment between national priorities and computing, and a focus on domestic supply chains and ecosystem development.
He also highlighted China's efforts to build full technology stacks, including hardware, systems integration and increasingly software. "The trajectory has been clear: sustained effort, growing maturity and expanding impact," he said.
Chinese researchers have made strong contributions in system design, parallel algorithms, numerical methods, large-scale AI and domain-specific applications, he added.
Despite geopolitical tensions, Dongarra said China could play a constructive role in global scientific collaboration, particularly in areas of shared benefit such as climate, energy systems, public health and sustainable computing.
"China should be at the table in global discussions on sustainable computing," he said. "The scale of deployment -- anywhere in the world -- will shape supply chains, standards and the environmental footprint of our field. If we can align incentives toward efficiency and reproducibility, everybody benefits." ■
