Table of Contents
Some of the biggest names in the tech sector came together this week to announce ‘Stargate’, a project they say will see $500 billion in investment for US-based artificial intelligence infrastructure. The joint venture, led by OpenAI, Oracle and SoftBank, aims to quickly build out colossal new data centers that will be crucial to future AI development. It will also support new power plants needed to power these notoriously energy-intensive AI models.
Stargate already has the blessing of newly inaugurated President Donald Trump, who said this week that he plans to “unleash” America’s fossil fuel industry. Loose regulations on oil and gas extraction will make fossil fuels the obvious and cheapest choice to achieve Stargate’s ambitious AI agenda. US AI companies, some of which have publicly committed to offsetting carbon emissions with green energy sources, may now be on the verge of doubling down on fossil fuels during a second administration. “Drill baby drill” is coming for AI.
What is Stargate?
Donald Trump discussed how Stargate could turn out during a press conference Tuesday evening. Standing next to him were three of the plan’s key architects: Larry Ellison, co-founder of Oracle, Sam Altman, CEO of OpenAI, and Masayoshi Son, CEO of SoftBank. In theory, the project aims to unlock $500 billion in funding to build new US-based data centers intended to power AI in the coming years. SoftBank is the main backer of the entity and OpenAI will be responsible for the project’s operations. Critics, including close Trump ally Elon Muskhave wondered whether the companies can actually secure the promised investment amount. Regardless, Stargate partners say they are already committing $100 billion to build out the first set of data centers in Texas. Writing in one this week’s blog postOpenAI said the multi-year effort could ultimately create hundreds of thousands of American jobs and “secure American leadership in AI.”
“I think this will be the most important project of this era,” Altman told Trump during Tuesday’s press conference. “We couldn’t do this without you, Mr. President.”
All these developments cannot come fast enough for AI companies. Companies, both in the US and abroad, are rushing to ship products and continually improve their AI capabilities. All this requires a wealth of data and servers that must be housed in energy-intensive data centers. A Goldman Sachs prediction Last year it was estimated that these types of data centers could account for 8 percent of total U.S. electricity consumption by 2026, nearly tripling from just 3 percent in 2022. To put that in perspective, the International Energy Agency (IEA) estimates The power demand of AI and cryptocurrency together will likely add “at least one Sweden or at most one Germany” to electricity demand next year. Those estimates were made before Trump’s victory in 2024. Now, with the president’s vocal support of Stargate, the total number of data centers and the electricity they require could rise even further.
AI’s insatiable energy hunger
Newer developments in generative AI are uniquely responsible for recent peak in electricity consumption. AI researcher Jesse Dodge told NPR last year he estimates that a single query to OpenAI’s ChatGPT could require as much energy as it takes to power a light bulb for 20 minutes. Queries sent to large language models (LLMs) were trained on trillions of data parameters reportedly require 10 times as much energy as a simple Google search. Energy consumption increases even further with more complex AI-generated content. Research has shown that a single image, created by an AI model such as DALL-E, requires about half of that the same amount of electricity required to fully charge a smartphone. AI-generated video requires even more electricity.
And even though major tech companies like Google and Apple have publicly committed to ambitious renewable energy policy goals, much of this new AI electricity demand is currently covered by coal and natural gas. Googling released a sustainability report last year they revealed that greenhouse gas emissions had risen 48 percent since 2019, a figure they attributed to “further integration[ing] AI in our products.” Microsoft, which supports OpenAI, published its own report they admit that greenhouse gas emissions have increased by 29 percent since 2020. They are not the only ones. Analysts at Goldman Sachs say the growing energy needs propelled by the broader AI race will lead to energy demand “unseen in a generation.”
Investments in nuclear energy and green energy will not save the day
Technology companies have made some efforts in recent years to satisfy their growing appetite for energy with renewable sources. These companies have done this together billions spent on wind and solar energy projects intended to offset their use of fossil fuels. The biggest players such as Microsoft, MetaAnd Amazon are also investing hundreds of millions of dollars in upgrading existing and creating entirely new nuclear power plants. Utilities are equal consider restarting decommissioned reactors to also take action. In theory, these could one day provide reliable and clean energy 24 hours a day to support AI energy while keeping CO2 emissions at bay.
But it will take years or even decades before new nuclear power plants are fully operational. Companies want the extra power now. The AI race won’t wait for nuclear power or even geothermal power to mature. Energy experts told it before Popular science that means dirty but undeniably cost-effective fossil fuels are the only solution that will fully realize AI’s rapid timeline. Goldman Sachs’ forecast noted previous estimates that fossil fuels could account for 60 percent of the new energy used to power data centers in the coming years.
“I don’t think the world is prepared for what’s going to happen in terms of AI demand,” said Jamie Beard, executive director of Project InnerSpace, a nonprofit focused on promoting geothermal energy. Popular science.
Trump’s message to AI: ‘Drill baby drill’
A new Trump administration is already making it easier to power data centers with fossil fuels. In the three days since he was sworn in, President Trump has done just that declared a so-called “national energy emergency” and determined to turn back key climate promises from the Biden administration. Trump has already done that There were black restrictions on oil drilling in Alaska, car emissions standards overturnedAnd has lifted a pause on approving applications for liquefied natural gas exports. Although legally questionableexperts say there is a unique energy “emergency” could give Trump the ability to expedite new permits for fossil fuel infrastructure, which could in turn increase energy supplies and reduce costs for data center owners. All of this, Trump said at his inauguration, is intended to “unleash” America’s energy sector despite some analysis suggesting that the world might be an oversupply of oil and natural gas is approaching.
“We will drill, baby, drill,” Trump said during his inauguration speech. “We will be a rich nation again, and it is the liquid gold under our feet that will help us get there.”
That’s likely all welcome news for data center owners and AI companies looking for cheaper, more readily available electricity. Although members of the Stargate Project have not publicly said where they plan to get most of their energy from, their close ties to the Trump administration and the sheer size of the new infrastructure they want to build make fossil fuels the obvious choice. horizontal choice. Oracle, OpenAI and SoftBank did not immediately respond to requests for comment. Trump said in a call with Stargate executives on Tuesday that his administration would make it easier for these companies to generate electricity and suggested they could even do so on-site in newly built “AI factories.”
“They need to produce a lot of electricity and we will make it possible for them to produce that very easily, in their own plants if they want to,” Trump said. “They are going to build power generation in the AI factory and that will be incredible.” It is unclear whether building energy activities in an AI data center is possible in practice.
An influx of fossil fuel-powered US AI companies threatens to increase carbon emissions, raise global temperatures and undermine the climate progress of recent years. Some of AI’s biggest proponents argue that this temporary toll on the environment could be worth further developing a technology that, when mature, can be used to combat the most extreme impacts of climate change. Some AI models are already getting used to it increase energy efficiency And improving air pollution monitoring. AI also plays an important role identifying areas rich in critical minerals needed for the production of electric vehicles and other lower-emission products. But all that risks coming at the cost of further warming the planet, which environmental scientists say contributes to hurricanes, wildfires and many of the other natural disasters that have become increasingly common in recent years. That could make the AI energy gamble look less like a logical gamble and more like a cynical stumble into an oncoming fire.
Win the holidays with PopSci’s gift guides