Abstract:

Amid the rise of computing-power synergy on the national agenda, grid enterprises offering external computing-power services must innovate their business and scheduling models. Adopting a “power-carbon-computing” perspective, this study proposes a composite-token-based business model, defines a four-dimensional attribute structure (computing specifications, green rights, migratability, and interruptibility), and formulates a multiplicative-factor pricing model. Full-chain costs and profits are compared on a per-GPU-hour basis across three models: whole-card leasing, pure computing-power tokens, and composite tokens. A token routing algorithm is then developed, weighting power price as the primary signal and renewable output as a secondary one. Customers receive discounts by relinquishing node-fixation or task-continuity rights, while the enterprise monetizes this flexibility through coordinated scheduling to preserve its margin. A GPU-as-a-Service (GPUaaS) model and a phased implementation roadmap are also presented, providing a reference for grid enterprises to develop differentiated computing-power operations.

Key words: computing-power synergy, composite token, power-carbon-computing integration, computing-power pricing, scheduling algorithm, renewable-energy consumption, new power system