Orient Securities Company Limited has released a research report stating that the synergy between computing power and electricity possesses immense potential in the era of rapid AI development. The significant increase in computing power creates substantial elasticity in electricity demand, which is expected to first enhance the profitability of power assets by improving regional electricity supply and demand dynamics. The firm constructed a three-dimensional measurement model of "Single Token Inference Energy Consumption × Daily Average Token Calls × PUE." Under a neutral scenario, it estimates that China's data center electricity consumption could reach 800 billion kilowatt-hours by 2030. The firm believes the long-term pricing logic for power assets may transition from "cost competition" to "value discovery." The main viewpoints of Orient Securities Company Limited are as follows:

Industry Volume: Computing Power Electricity Consumption Has Room to Exceed Expectations; Supply and Demand in Key Computing Hubs May Tighten First

1) There is some divergence in the market regarding the long-term growth space for electricity consumption driven by computing power. Using the aforementioned "Single Token Inference Energy Consumption × Daily Average Token Calls × PUE" model, the firm estimates China's data center electricity consumption could reach 800 billion kilowatt-hours by 2030 under a neutral scenario. Further analysis of the three core variables reveals that daily average token calls, as a demand-side driver, have a growth boundary that has not yet converged and virtually no physical upper limit. This factor is the dominant variable determining the elasticity of computing power's electricity consumption. This suggests the long-term ceiling for computing power electricity use may be higher than current market expectations, making power supply assurance a core competitive variable in the computing power era.

2) From a supply-demand perspective, the high growth of new energy installations at the national level is sufficient to cover the incremental electricity demand from computing power, making systemic "power shortage" risks relatively low. However, new computing capacity is concentrating in green electricity-rich regions like Inner Mongolia, Ningxia, Gansu, and Guizhou. The injection of this high-certainty electricity demand significantly alleviates the curtailment dilemma for new energy, shifting it from "passive curtailment" to "active matching."

Industry Price: Underestimated Electricity Price Elasticity and Unpriced Token Sharing Option

1) The market is generally pessimistic about the profit prospects of computing-electricity synergy projects. The firm believes the current industry pain points are temporary. In the future, as falling GPU prices increase the weight of electricity costs and the power supply-demand balance in key computing hubs shifts from loose to tight, the pricing logic for green electricity is expected to revert from the passive situation of "curtailment suppressing prices" to market-based pricing that reflects resource scarcity. An upward shift in the electricity price center will significantly release project profit elasticity.

2) Against the backdrop of major tech companies outsourcing capital expenditures, power generation enterprises with access to high-quality green electricity resources are natural recipients of this Capex outsourcing, possessing both the qualifications to secure projects and leverage in sharing negotiations. The profit model may transition from simply "selling electricity" to "token sharing." The profit ceiling for power operators will be systematically raised, and their valuation logic may shift from traditional utility attributes to that of a growth stock.

Investment Recommendations and Targets

Computing-electricity synergy is expected to drive a revaluation of power asset value. In terms of volume, the high growth of computing power brings massive electricity demand, providing stable and ample consumption space for green power. In terms of price, as the weight of electricity costs rises and power supply-demand in key computing hubs tightens, green power pricing may revert from "curtailment suppressing prices" to reflecting its market value based on resource scarcity. In the long term, the profit model may transition from traditional "electricity sales" to "token sharing." The profit ceiling and valuation center for power operators are expected to rise simultaneously.

Risk Warnings

Uncertainty regarding the progress of electricity market reform; intensifying industry competition; AI demand falling short of expectations; policy implementation lagging behind expectations; changes in assumptions affecting calculation results.