Computing power underpins the advancement of disruptive technologies such as artificial intelligence and blockchain. Their rapid development has driven an exponential increase in global demand for compute, prompting countries to accelerate the construction of related infrastructure. Yet this surge has also brought to the fore a series of ethical concerns, including tensions between expanding computing capacity and the carrying limits of ecosystems, as well as disparities between technologically advanced and less advantaged nations. In this context, it is imperative to build a sustainability-oriented ethical framework for computing power.

Efficiency, low-carbon, fairness

The development of computing capacity brings with it mutil-facted challenges. Environmentally, the extraction of rare metals used in chip manufacturing can damage fragile ecosystems, while electronic waste poses long-term risks to soil and water safety. In terms of resource allocation, countries with advanced technologies risk widening the global computing power divide through technological monopolies and export controls, while tech giants leverage capital advantages to squeeze the living space of small and medium-sized enterprises (SMEs). When breakthroughs in core technologies stall, the industry often resorts to stacking energy-intensive hardware to boost performance. In light of these intersecting pressures, a sustainable ethical framework—anchored in efficiency, low carbon, and fairness—is needed to balance technological innovation, ecological protection, and social equity.

The principle of efficiency calls for minimizing the computing power required for a given task through algorithmic optimization and architectural innovation. Its ethical value lies in shifting the impetus for technological progress away from sheer hardware expansion toward innovation-driven improvements in energy efficiency, helping steer the industry toward more sustainable growth. Over the past decade, technological approaches that failed to substantially enhance energy efficiency have largely been phased out; competition has moved from expansive “chip stacking” to a more discerning “energy-efficiency race.”

The low-carbon principle requires embedding ecological thresholds into planning for computing power development and establishing long-term, multi-dimensional constraint mechanisms across generations, regions, and technological domains. Regional deployment should be aligned with ecological carrying capacity to prevent technologically advanced areas from offloading environmental costs onto ecologically vulnerable regions. Technological upgrading should emphasize synergistic improvements in algorithmic and hardware energy efficiency, breaking the linear relationship between computing power growth and energy consumption.

The fairness principle seeks to mitigate monopolies over computing power and narrow the global divide through institutional innovation and technological breakthroughs, promoting broader access to computing resources. Establishing reasonable pricing mechanisms and regulatory frameworks can curb market concentration, while the adoption of distributed technologies helps reduce resource centralization, enabling SMEs and individual users to access compute at lower cost. Shared platforms for computing power, together with cloud computing and pooled resource architectures, will help achieve flexible, efficient resource allocation, reducing waste and preventing idle capacity.

Pathways for implementation

The principles for sustainability-oriented ethics of computing power can gradually be implemented through coordinated efforts in policy formulation, ethical design, social participation, and international cooperation.

First, governments should steer the development of computing power to ensure alignment with environmental protection, resource equity, and social justice. For example, China’s Action Plan for the High-Quality Development of Computing Power Infrastructure (2023) is aimed at promoting the construction of a nationwide integrated computing power network, enhancing computing facility utilization efficiency, and facilitating regional coordination in computing power development.

Second, ethical principles should be integrated into hardware design and software optimization from the earliest stages of R&D. Approaches such as Value-Sensitive Design and Responsible Innovation can embed ethical principles throughout a technology’s life cycle, minimizing value conflicts and ethical risks. A dedicated ethical assessment mechanism should be established to dynamically monitor and evaluate R&D and application processes, identifying and mitigating emerging risks that may provoke ethical controversies.

Third, joint participation of large enterprises, research institutions, and the public forms the social foundation for advancing computing power ethics. For instance, CFFF, a cloud-based intelligent computing platform launched in 2023 by Fudan University, Alibaba Cloud, and China Telecom, offers ultra-large-scale parallel intelligent computing through an advanced public cloud model and supports the training of large models with hundreds of billions of parameters.

Fourth, a global standard for computing power ethics should be developed, combining technology sharing with ethical co-construction. The construction of cross-border computing power platforms and regional distribution centers can help optimize resource allocation and dynamically distribute resources. Developed countries can support infrastructure development in developing countries through technology transfer and computing power assistance.

Li Lun is a professor from the School of Humanities at Dalian University of Technology.