By 2027 the US will need up 60 gigawatts to power roughly 3,000 data centers, and it has enough natural gas to do so, but ESG policies and logistics are preventing that from happening, according to a Goldman Sachs executive at the Doha Forum in Qatar.
Speaking at the Doha Forum in Qatar during a session on “The Geopolitics of Artificial Intelligence” over the weekend, Goldman Sachs president of global affairs Jared Cohen said that in order to meet AI demands, the US will need to step-up its energy output for data centers from 17 gigawatts last year to another 35 to 60 gigawatts of baseload power by 2027.
“For these AI workloads, the data centers that are required to power them are ultra-high-density, and they require a single concentrated source of power,” said Cohen.
“Intermittent power like wind and solar can’t be used for it; it requires 24/7 power. What it means is you need baseload power, so you think coal, nuclear, natural gas,” he added.
Cohen said that the US had enough natural gas at the present to meet the AI demand of 35 to 60 gigawatts of power, but that money, logistics, and ESG policies were preventing that from happening.
“The US actually has enough natural gas to do this, but it’s in North Dakota and it’s in all these other places, but it’s prohibitively expensive to transport it to where the data centers need to be built or are located, and it’s prohibitively fraught because it has to go through all of these different political jurisdictions that are pushing back for ESG reasons,” he said.
Recognizing that renewables alone can’t power AI data centers, World Economic Forum (WEF) president Børge Brende told the same panel that wind and solar should still be deployed, and that natural gas, nuclear, and hydro should be used as backups.
“You have to have access to competitive, cheap electricity. Hopefully, I think it should be renewables, but you have to have a backup being hydro, nuclear will take time, but of course with gas […] is of course a lot more preferred than if it’s coal,” said Brende.
The estimated energy demand for AI data centers increases with every passing month.
Speaking at the WEF’s Special Meeting on Global Collaboration, Growth and Energy Development in Riyadh, Saudi Arabia last April, Fink told the panel on “Investing Amid Global Fracture” that intermittent wind and solar were not going to cut it when it came to powering AI data centers.
Eight months after Fink’s remarks that 30 gigawatts would be required by 2030, that figure has now jumped to between 35 and 60 gigawatts in a shorter time frame, by 2027.
While unelected globalists push expensive, unreliable renewables like wind and solar on the rest of the population in the name of saving the planet and net-zero policies, the elites are saying that AI data centers will require enough energy to power dozens of cities.
Image Source: Screenshot of Goldman Sachs President of Global Affairs Jared Cohen, Doha Forum YouTube