OpenAI anticipates rapid data center scaling thanks to Trump

LAS VEGAS — OpenAI expects that President Donald Trump’s policies are going to aid the artificial intelligence software company by helping build out new data centers, according to one company leader.

Speaking at the inaugural HumanX Conference in Las Vegas, OpenAI Chief Product Officer Kevin Weil said that as OpenAI continues to refine its products to be smarter, plans to build more U.S. data centers tailored to process the large volumes of training data required will be critical. 

“One of the things I'm really excited about in the current administration is: Trump is going to help us build a lot of infrastructure,” Weil said. “We need to build a lot of infrastructure. What we've seen over the past 10 years is every ounce of compute that we can apply to a model makes that model smarter. When the models are smarter, you can solve the problems with them, they're more valuable for people.”

Weil’s comments follow a January announcement from several members of industry of a $500 billion dollar investment to create more AI infrastructure, known as the Stargate Project. Unveiled at the White House alongside Trump, the project’s first buildout is occurring now in Texas. OpenAI CEO Sam Altman and leadership from partnering companies SoftBank, Oracle and MGX confirmed the project's groundbreaking.

The infrastructure will provide OpenAI’s products more access to compute power for enhanced learning and data processing. Weil said that he anticipates a quick scale-up of these data centers thanks to the Trump administration’s expedition of the requisite permitting and restrictions.

“More compute will make our models smarter, [and] smarter models will help us build better products for you,” he said. “And to do that, you've got to build a lot, and you need to get faster permitting and all these things. I think the administration is going to help a lot with that, which is great.”

Weil noted that the company is still prioritizing user safety despite its enthusiasm to deliver faster, more accurate products at a cheaper cost and the broader Trump administration optimism towards unlocking AI’s potential. He confirmed that OpenAI still employs red teaming efforts to search for software errors and consults external expertise. 

“We want to build faster, and we want to do it the right way,” he said. 

Weil’s comments touch on several central concerns within the AI industry. AI companies around the world were shocked by reports of Chinese-developed AI software DeepSeek rivaling top AI models’ output capabilities while demanding less compute resources. Despite the lack of transparency around DeepSeek’s infrastructure demands, the model underscored the desire Silicon Valley and Capitol Hill share: ensuring the U.S. leads the way in advanced AI development.

Weil described the cost of intelligence — or the cost it takes to train a given AI model to produce highly accurate results — as declining annually. He attributed this to a convergence of computing components, like GPUs and the AI models themselves, that are getting smaller and faster. 

“One of the bets that we consistently make is that that is going to continue,” Weil said regarding the efficiency of the entire AI industry. “And one of the ways that we bring that to life is whenever we have an opportunity to bring down the cost of our products, we do.”