The race for AI dominance is in full swing. OpenAI and ChatGPT may have opened the eyes of the world to AI, but make no mistake: Big Tech companies are all engaged in the competition for artificial intelligence superiority. As a result, companies like Google, Meta, Microsoft and Amazon are investing billions in these new technologies, and investments are pouring into new data centers to drive AI advancement. These efforts may seem admirable in many ways, but the electricity demand of AI data centers is tremendous and are presenting additional problems. Specifically, environmental risks and resource shortages could be a real issue. And though Big Tech companies play these concerns down, experts suggest otherwise.
(Would you work for an AI boss? Read this Bold story and decide.)
The dilemma for these companies and for the world at large is the challenge of converting to green energy while fueling AI growth. One the one hand, Big Tech companies have committed to green energy and no carbon emissions by 2030. On the other, the competition for artificial intelligence and its positive impacts on the world undermine these efforts. Perhaps AI can make major contributions that might reduce carbon emissions and fuel energy innovations. But science suggests otherwise, given the massive electricity demands of AI data centers being built today. Everyone agrees, however, that a major breakthrough is needed if a happy balance is to be achieved. It’s the probability of such a breakthrough that stakeholders seem to disagree about.
In the Midst of a Data Center Boom
In effect, all the major players in the competition for artificial intelligence knows data is the key. To train AI and to realize its potential, large amounts of data and data power are required. Because of this, the nation is going through a boom in data center construction. No existing warehouse is safe at present or large parcels of land. That is if they are located near ample water supply and can meet the electricity demands of AI data centers. In total, there are more than 2,800 data centers in the U.S. with many more planned in the near future, and if they’re being used for AI, their electricity requirements are profound. It has been estimated that these data centers will consumer 8% of the nation’s total electricity by 2030. This represents three times the current percentage that these centers use. And it’s the equivalent of adding nearly 16 million gas-powered cars to U.S. highways.
All the Big Tech companies are participating in this data center boom. The competition for artificial intelligence has encouraged each to invest heavily in these new projects. Meta recently built a major data center complex in Iowa that equates to running 7 million laptops daily for eight hours constantly. Microsoft has invested in a $3.3 billion data center in Milwaukee. And in Salt Lake City, both Meta and Google have major data center complexes planned. Google’s will take up 300 acres of land while Meta’s will cost $1.5 billion. Meta’s in particular will reflects the growing electricity demands of AI data centers. Its new complex will require the equivalent power needs of that supplied by a single nuclear reactor. Given this, data centers involved in the competition for artificial intelligence will also be competing with cities for power. It seems inevitable that something will have to give.
Big Tech’s Proposed Energy Solutions
It’s interesting that Microsoft, Meta, Google and Amazon have each committed to a 2030 date for being emission-free. But given data center growth in the competition for artificial intelligence, this seems unlikely. Indeed, each of these companies are purchasing and investing in green energy technologies with their data centers. Wind, solar, and geothermal sources of energy are bought to help support the electricity demands of AI data centers. But these energy credits routinely fall quite short of data center needs. This has resulted in several coal plants across the country being kept open that were scheduled to close. And it’s resulting in cities investing more heavily in natural gas for their future. While green energies will help, they alone cannot support AI growth. Many expect natural gas to be the major energy supporter of AI data centers in the near future.
Big Tech companies have alternative plans. For example, Microsoft has partnered with Helion, a nuclear fusion startup company. Helion believes it can harness nuclear fusion energy by 2030 to meet the electricity demands of AI data centers. They believe this is possible by using a magnetized vacuum chamber. But such solutions have eluded some of the world’s top scientists for decades, making such a timetable unlikely. Google is looking into geothermal as a potential solution as the competition for artificial intelligence heats up. It is working with Fervo Energy in Utah and Nevada for these purposes. This requires boring into the earth’s crust 10,000 feet to convert heat into electricity. And finally, some are exploring connecting small nuclear reactors to their data center campuses. Meta and Amazon seem the most likely candidates for these approaches. But in each case, feasibility issues may prevent anticipated results. It remains to be seen if Big Tech can pull off a miracle in this regard.
A Conflict of Interests
At a recent conference, OpenAI’s chairman, Sam Altman, admitted the competition for artificial intelligence required an energy breakthrough. Thus, there’s little debate about the anticipated electricity demands of AI data centers in the near future. There is, however, opposing views on what needs to happen. Environmentalists and even many smaller cities see data centers as a major threat. They therefore are concerned about AI data center growth and believe it should be curbed. But Big Tech and AI startups argue the opposite. They believe through AI advancements, solutions to energy needs and climate change can be better realized. Such solutions include improved management of power grids and new ideas concerning green energy technologies. It’s not clear if this is indeed true, but there’s no question electricity demands of AI data centers are concerning. Unless new and clean sources of “juice” are found, something will have to give.
