Just a week into the new Trump administration, the AI sector has already witnessed its first intellectual battle. The clash arises from two important news: the announcement of the Stargate project, which intends to spend US$500 billion—more than the Apollo space program—on new AI data centers, and the launch of a powerful model from China. Together, this information raises crucial questions about how necessary the rush for more data centers—with its heavy environmental impact—is.
Recalling the first point: OpenAI, Oracle, SoftBank and an Abu Dhabi-based investment fund called MGX plan to invest up to US$500 billion in building gigantic data centers in the United States to improve AI. Much of the groundwork for this project was laid in 2024, when OpenAI increased its lobbying spending sevenfold (as first reported last week) and AI companies began pushing for policies less focused on controlling problems like deepfakes and misinformation, and more on getting more power.
Still, Trump took credit for the project from technology leaders when he announced it on his second day in office. “I believe this will be the most important project of this era,” OpenAI’s Sam Altman said at the launch event, adding, “We couldn’t do it without you, Mr. President.”
That’s an incredible amount, just a little less than the inflation-adjusted cost of building the U.S. highway system over more than 30 years. However, not everyone sees Stargate as a public benefit. Environmental groups warn that it could overload local power grids and further increase energy costs for the rest of the population, which does not consume electricity to train and deploy AI models. Previous research has also indicated that data centers tend to be built in locations that use much more carbon-intensive energy sources, such as coal, than the national average. It is unclear to what extent, if at all, the Stargate will rely on renewable energy.
The project’s most vocal critics, however, include Elon Musk. None of Musk’s companies are involved in Stargate, and he has publicly tried to sow doubt about whether OpenAI and SoftBank actually have the resources needed to bring the plan to fruition, a claim that Altman refuted on Platform
Now, the second news. On the day of Trump’s inauguration, a Chinese startup launched an AI model that left many influential people in Silicon Valley worried about the competition. (This close timing was certainly no coincidence.)
The system, called DeepSeek R1, is a reasoning model. These cognitive structures are designed to excel at math, logic, pattern identification, and decision making. DeepSeek has been shown to be able to “reason” about complex problems as well as one of OpenAI’s reasoning models, o1—and more efficiently. Furthermore, DeepSeek is not a top-secret project kept under lock and key like OpenAI. It was made publicly available for everyone to see.
DeepSeek was launched despite the US’s top priority: beating China in the AI race. This goal was one of the main drivers behind the CHIPS Act of 2022, created to boost domestic semiconductor production. It also influenced the stance of technology companies such as OpenAI, which began collaborating with the national security sector and partnered with Anduril, a defense technology company, to help the military shoot down drones. Additionally, it has led to the imposition of export controls that limit what types of chips Nvidia can sell to China.
DeepSeek’s success indicates that these efforts are not working as well as U.S. AI leaders would like. (Although it is important to note that the impact of export controls on chips is only felt after a few years, so this policy would not be expected to impede a model like DeepSeek at this time).
Still, the system poses a threat to some technology giants. Why pay for an expensive OpenAI model when DeepSeek is free? According to information from the portal The Information, even other companies that produce open source models, especially Meta, are panicking about the new competition. Meta set up several “war rooms” to understand how DeepSeek managed to be so efficient. (Just days after announcing the Stargate project, Meta declared that it would increase its investments in AI infrastructure by 70%.)
What does all this mean for the Stargate project? Let’s look at why OpenAI and its partners are willing to invest $500 billion in data centers. They believe that artificial intelligence, in its various forms—not just chatbots, generative videos, or new AI agents, but also unrevealed developments—will be the most profitable tool ever created by humanity. Furthermore, they believe that access to powerful chips inside gigantic data centers is the key to achieving this goal.
DeepSeek, however, defied this logic. He wasn’t trained on unreleased chips light years ahead of the competition. As far as we know, it did not need the astronomical amounts of computational power and energy that models from North American companies usually demand. Its creators made strategic decisions focused on efficiency.
Theoretically, this could make the Stargate project less urgent and necessary. If, by analyzing DeepSeek, AI companies learn new ways to optimize the use of existing resources, perhaps building massive data centers will no longer be the only winning strategy for advancing AI. This would be a relief to many communities affected by the impacts of data centers, such as high carbon emissions, wasted drinking water for cooling, and strain on local electrical grids.
So far, however, DeepSeek doesn’t appear to have prompted a change in this kind of thinking. OpenAI researcher Noam Brown wrote on Platform X: “I have no doubt that, with even more computing power, it [the company’s artificial intelligence] would be even more powerful.”
If this logic prevails, then companies with the most computing power will continue to dominate the industry—and achieving that power appears to be worth at least $500 billion to the AI giants. But it’s worth remembering: announcing the investment is the easiest part.
( fontes: MIT Technology Review)
