BloombergNEF's Izadi-Najafabadi delivers keynote at Energy Leaders Summit
“Competition for data center locations is intensifying: grid and land permits determine investment”
“Spending alone is unlikely to secure competitiveness in an AI model market led by the US and China”
South Korea needs to improve the reliability of its power supply and data center permitting conditions while developing independent AI model capabilities to avoid falling behind in global competition in the artificial intelligence industry, a speaker has recommended.
Ali Izadi-Najafabadi, head of global transition analysis at BloombergNEF, emphasized these points in a keynote address at the Energy Leaders Summit on November 5, identifying AI and energy as two forces reshaping the global economy.
According to Izadi-Najafabadi, global data center capacity has tripled over the past 6 years. Data centers for AI training are larger than existing facilities and have higher server-rack power densities, driving a steep increase in electricity demand. Izadi-Najafabadi forecast that annual spending on data center construction would reach $1 trillion in 2026, with cumulative investment reaching about $7 trillion by 2031.
Izadi-Najafabadi identified five factors determining data center locations: access to power, land and permitting, taxation, access to fiber-optic networks, and the clustering benefits of existing data centers. In BloombergNEF's assessment, Virginia in the United States ranked highest in competitiveness despite difficulties connecting to the grid. In the Asia-Pacific region, Indonesia received a relatively favorable assessment because of its available generation capacity and permitting conditions.
By contrast, South Korea still has areas to improve, Izadi-Najafabadi said. “South Korea faces difficulties securing power, and land permitting is also more demanding than in some competing countries,” Izadi-Najafabadi said. “Policy responses are needed in areas beyond energy as well.”
Izadi-Najafabadi also explained the impact of data center expansion on power generation markets. While it is stimulating investment in gas-fired generation in the United States, renewables and batteries are expected to play larger roles elsewhere. However, without policy responses, rising demand could extend the operating lives of existing fossil-fuel power plants, Izadi-Najafabadi cautioned. Malaysia is reviewing plans to retire some coal-fired power plants because of surging electricity demand, Izadi-Najafabadi said.
Izadi-Najafabadi also assessed that simply allocating substantial budgets would not make it easy to secure competitiveness in the AI market. Leading companies such as OpenAI and Anthropic are already generating substantial revenue, with some exceeding their own forecasts, but that revenue is still insufficient to cover the expenditure required to build data centers and develop cutting-edge models.
In a market where recovering investment costs is this difficult, South Korea would struggle to narrow the gap with the United States and China if it approached frontier-model development solely in terms of immediate costs, Izadi-Najafabadi explained.
“If left entirely to market forces, existing developers in the United States and China are likely to dominate the AI model market,” Izadi-Najafabadi said. “Middle powers such as South Korea risk being unable to secure their own frontier-model development capabilities.”
Izadi-Najafabadi emphasized that although energy accounts for a small share of the total cost of AI model development, power supply is a prerequisite. Continuing AI development requires relieving grid congestion and securing sufficient supply capacity, Izadi-Najafabadi explained.
Original Korean article: Electric Times
English brief by Korea Wind Intelligence, koreawindintel.com
