Hello?
These days, the hot topic in AI data centers is the atmosphere of expansion due to memory supply and shortage of power and infrastructure. In particular, the amount of power consumed by AI is not just a matter of 'needing a lot,' but there is also a growing trend of opposition from residents in DC locations.
In response, leading countries that are pioneering the adoption of AIDCs, including the United States, have approved or are considering new introductions of nuclear power plants and SMRs.
However, since nuclear power plants and SMRs cannot be built and operated as quickly as thermal and wind power (within 2~3 years), it is a foregone conclusion that there will be a power shortage by 2030 at the current time of late 2026.
The table below summarizes the content of a domestic securities report. Of course, it is necessary to consider that future demand forecasts are estimates.
Simply looking at the global AI demand forecast, there will be a shortage of about 20GW in AI DC power demand by 2030. To put this into a very simple and naive example, this is equivalent to about 10~15 nuclear power plants.
Of course, the GW unit used here is different from GWh, and it is not necessary to have a constant low capacity. However, even if we start building nuclear power plants and SMRs now, it will be difficult to operate them by 2030. Therefore, some alternative will be needed between 27~29.
Among the alternatives, ESS will also be considered, as well as various methods such as solar power generation, gas turbine power generation, and extension of the operating life of existing nuclear power plants.
Therefore, for at least the next 2 years, it may be a good idea to pay attention to the stocks listed above as AI transitions from the learning stage to the inference stage.
I recommend reading the three recent reports attached for those who are interested.
