Diving overview:
Despite growing concerns about the impact of artificial intelligence on electricity demand and fossil fuel generation, NVIDIA CEO Jensen Huang said Friday that AI can help achieve global decarbonization.
AI initially requires a lot of energy to train, but in the long run it is more efficient than traditional computing, Huang said. He also argued that AI could be integrated into the power grid to deliver electricity more efficiently.
U.S. data center load is expected to increase from 19GW in 2023 to nearly 21GW this year, according to a Federal Energy Regulatory Commission report this month. Data center power demand across the United States is expected to increase to 35 GW by the end of this decade, according to a FERC report. Huang said this growth should be seen as a sign of prosperity.
Dive Insight:
Fan said in a webinar hosted by the Bipartisan Policy Center on Friday that AI should be seen as a potential climate change solution.
“Using artificial intelligence to solve problems consumes less energy than using computation to solve problems,” Huang said.
For example, he said AI models can predict the weather 1,000 times more efficiently than traditional computing. Training AI requires considerable energy upfront, but once the model is created, it can be used repeatedly around the world, ultimately leading to energy savings compared to traditional computing processes, he said. said.
AI could be used to upgrade the power grid, “redirect energy in smarter ways,” and perhaps improve carbon capture and sequestration technology, Huang said. . However, he acknowledged that the efficiency gains created by AI will likely be absorbed as society assigns more tasks to AI.
“Society will use the energy saved for other things. We call it prosperity and economic growth and improving quality of life. That’s what we want,” Huang said. spoke. “As a matter of fact, in absolute terms, I hope that someday people on the planet will consume more energy, because it is directly related to the quality of life. I hope you enjoy Mogako’s quality of life.”
EPRI estimated in May that data centers could consume 9% of U.S. electricity generation by 2030, double their current consumption.
Huang also expressed support for the recent agreement between Constellation Energy and Microsoft to restart the Three Mile Island nuclear power plant, saying the world needs all available energy sources. He said he was deaf.
“No single energy source is enough for the world,” he said.
But Huang said the U.S. should focus on building data centers in areas that are already energy-rich, rather than building new power grids to bring power to areas with high demand. AI can also be thought of as a type of energy storage, he said.
“Instead of transporting energy to where it’s needed, let’s transport data centers to where the energy sources are,” he said. “You can build a data center anywhere. Computers don’t care where you go to school.”
