Source: Finance Association
① In recent months, the claim that “the end of AI is electricity” has sparked a buzz in the industry. ② Arm CEO Rene Haas also said on Tuesday that if efficiency is not improved, AI data centers may consume 20% to 25% of the electricity demand in the US by 2030.
chip design company$Arm Holdings (ARM.US)$It has gained fame by designing various methods to minimize smartphone power consumption and extend battery life. Now, the company's head says artificial intelligence applications may also require the same changes to improve energy efficiency.
Arm CEO Rene Haas said in an interview with the media on Tuesday local time that if efficiency is not improved, the electricity consumption of artificial intelligence data centers may be as high as 20% to 25% of the US electricity demand by 2030. Today, that share is probably only 4% or less.
“Honestly, this is hard to keep going,” Haas said.
In fact, in recent months, claims that “the end of AI is electricity” have sparked a buzz in the industry. Tesla CEO Musk said in February of this year that the restrictions on artificial intelligence computation are predictable.
Musk said at the time, “I predicted a shortage of chips over a year ago; the next shortage will be electricity. I don't think there will be enough electricity to run all the chips next year”.
OpenAI CEO Altman also pointed out at the annual meeting of the World Economic Forum in Davos that artificial intelligence will consume more electricity than people expected, and future development will require energy breakthroughs.
These remarks or concerns have already contributed to a sharp rise in the share prices of power stocks and energy stocks in the US stock market.
The International Energy Agency (IEA) said in a January report that ChatGPT consumes an average of 2.9 watt-hours of electricity in response to a request — equivalent to lighting a 60-watt light bulb for just under three minutes, which is almost ten times the average power consumption of a Google search query.
The agency said that between 2023 and 2026, the AI industry's electricity demand is expected to increase more than tenfold.
In response, Haas stated, “If the power demand for these large data centers continues to increase, it will be difficult to accelerate the relevant technological breakthroughs we need.”
It is worth mentioning that on Tuesday, the US and Japan just announced a $110 million plan to fund AI research at universities in both countries. Arm and its parent company SoftBank Group will also jointly fund the program with $25 million. Arm's funding will be used for collaboration between Carnegie Mellon University in Pittsburgh, USA, and Keio University in Japan.
Haas said he hopes that the US-Japan research cooperation program mentioned above will focus on developing solutions to these electricity problems.
editor/tolk