To The Moon

The rapid advancement of artificial intelligence is generating unprecedented electricity demand, pushing global power grids toward an inevitable expansion and upgrade. At the G20 summit on September 1, Elon Musk issued a stark warning that the AI industry is facing a significant power supply shortage, projecting a minimum shortfall of 15 gigawatts for AI chips by 2027. This prediction stems from an expected 40% to 50% annual growth rate in AI chip production, which starkly contrasts with an electricity supply growth rate of only 10% to 20% per year in regions outside of China. Furthermore, he highlighted that companies like Google and Anthropic are renting computing power from SpaceX precisely because SpaceX has invested in its own power generation facilities.

Musk's concerns are far from unfounded, as recent data suggests that electricity supply has become the primary bottleneck limiting US AI computing power expansion. The US Energy Information Administration (EIA) clearly stated in its latest Short-Term Energy Outlook (STEO) that following a second consecutive year of record-breaking US electricity consumption in 2025, nationwide power use is projected to hit new all-time highs in both 2026 and 2027. This surge is driven by both explosive AI computing demand and the broader electrification of society.

On one hand, data centers are pushing US electricity consumption toward historic levels. According to data from the Electric Power Research Institute (EPRI), data centers currently account for 5% of US electricity demand, a figure that could triple by 2035. In Virginia, this share has already surpassed 25%. On the other hand, the pace of data center construction is significantly lagging behind planning due to power shortages. A research report from Data Center Watch revealed that in the first quarter of this year alone, the total value of data center projects blocked or delayed across the US reached approximately $130 billion.

JPMorgan reported that satellite image analysis indicates more than 60% of data center projects slated for completion in 2027 have not yet broken ground, and a further 7% are already experiencing delays, primarily due to power supply bottlenecks and public opposition. Amid this power supply-demand imbalance, grid construction has emerged as a global priority. Some analysts believe the main pressure driving up US electricity costs comes from transmission and distribution, rather than power generation itself. Brian Janous, a former Microsoft energy vice president, compares ultra-high-voltage transmission lines to the "interstate highways" of the grid system, noting that just as interstate highways enabled large-scale freight transport, the grid requires high-voltage backbone lines to support expanding electricity demand.

Greg Abel, often referred to as Warren Buffett's successor, has pointed out that the massive construction of data centers and the resulting pressure on the grid are creating significant growth opportunities for the utility sector. He projects that data centers' share of electricity consumption could rise by another 5 to 10 percentage points over the next five years, potentially reaching as high as 50% in the long term, making it the most promising growth curve in the energy business. The US grid is undergoing an epic expansion, with the country's three major regional grid operators (in Texas, the Mid-Atlantic, and the Midwest) having approved a combined $75 billion in transmission expansion projects earlier this year. Central to these projects is the construction of 765-kilovolt ultra-high-voltage lines, the highest operating voltage in the US, capable of transmitting six times the power of traditional lines. This "electricity highway" is set to expand to a total length of 10,000 miles, four times the current mileage of approximately 2,000 miles.

In South Korea, the largest power utility, Korea Electric Power Corporation (KEPCO), has proposed a prepayment plan to Samsung Electronics and SK Hynix totaling 25 trillion won (approximately RMB 123.8 billion) for electricity bills covering the 2027-2031 period, using the funds to build grid infrastructure for the Yongin and South Chungcheong semiconductor clusters.

The continuous expansion of AI data centers makes the rigid growth of computing power electricity demand an irreversible trend. The increasing power density and large load fluctuations of data centers will pose severe challenges to grid stability, making "Computing-Power Synergy" an essential path to address these issues and support the high-quality development of the computing industry. Recently, the National Energy Administration held a working meeting on new grid construction, deploying major tasks for building a new type of grid. On August 19, the National Development and Reform Commission convened a coordination meeting for the "Six Networks" major projects to establish a "2+3+N" coordination mechanism involving computing networks, new grids, and next-generation communication networks. On August 3, the National Development and Reform Commission and the National Energy Administration issued the "15th Five-Year Plan for New Power System Construction," proposing the initial establishment of a safe, reliable, green, low-carbon, resilient, intelligent, and flexible new grid.

Orient Securities stated that the long-term ceiling for computing power electricity demand may be higher than current market expectations, and power supply capability is becoming a core competitive variable in the computing era. Computing-power synergy is expected to drive a revaluation of power assets. From a volume perspective, the surge in computing power creates massive electricity demand, providing a stable and ample absorption space for green power. From a price perspective, as the weight of electricity costs rises and power supply in computing hubs tightens, green power pricing may shift from "curtailment-induced low pricing" to reflecting the scarcity value of resources in the market. The firm believes that in the long run, the profitability ceiling and valuation center for power operators are expected to rise simultaneously, with business models potentially transitioning from traditional "electricity sales" to "Token revenue sharing."

Dongguan Securities noted that computing-power synergy deeply integrates computing load with green power supply, providing data centers with stable, low-cost green electricity while improving renewable energy consumption efficiency, achieving a win-win for both supply and demand. CITIC Securities research suggests that computing-power synergy is the key infrastructure for expanding domestic computing power, and without it, there can be no high-quality development of domestic computing. China's energy system provides vast expandable space for domestic computing power, free from grid connection queuing constraints, allowing each unit of CAPEX to be more efficiently converted into actual computing capacity.

The National Energy Administration projects that China's data center electricity consumption will reach 800 billion kWh by 2030, and based on this guidance, analysts predict an incremental data center scale of 78.5GW during the "15th Five-Year Plan" period, corresponding to capital expenditure of RMB 10.7 trillion. Computing-power synergy, divided into both inside and outside data center parks as the primary means of coupling electricity and computing power, is projected to require a total investment of approximately RMB 2 trillion.