The Oval Office ‘Stargate Project’ reveal was just more tech industry genuflecting

Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. You can sign up to receive this newsletter every week here.

Trump’s ‘Stargate Project’ reveal had little to do with reality

“Stargate” sounds like a movie, and where Donald Trump is concerned it is indeed mainly a performance.

Stargate, a new joint venture formed by OpenAI, Oracle, and SoftBank, aims to build up to $500 billion worth of new data centers over the next four years to run large AI models. OpenAI’s Sam Altman, SoftBank’s Masayoshi Son, and Oracle CEO Larry Elder joined Trump in the Oval Office Tuesday to outline the new project. Trump took to the podium to hitch his administration to the job creation and geopolitical rewards of the current AI boom.

The announcement in the Oval was choreographed to look like a public-private partnership, but it’s not. Trump pledged no federal federal funding or tax breaks; he only talked vaguely about using “emergency declarations” to smooth Stargate’s way. (In theory the White House could work with Texas lawmakers to bypass regulatory land-use hurdles, etc.). 

“This is very much a private project—it’s no different than a big company deciding to [build] their factory in the U.S., and so that’s good for America in the sense that it creates jobs,” says Abhishek Nagaraj, a professor at the University of California–Berkeley’s Haas business school. “This isn’t necessarily going to do all the things that we’ve been asking the government to do in terms of supporting AI innovation, research, etc. within the U.S.” 

The initiative was already well underway before Trump’s second term began. Altman has been talking for months about raising trillions to build more data centers for his company’s next generation of compute-hungry models. OpenAI and Oracle had already been working together on those data centers, and have in fact broken ground on 10 of them, Oracle’s Ellison said. SoftBank has been trying to invest in OpenAI since 2023, finally getting its chance last November when it put $1.3 billion into the company. Son pledged to invest $100 billion back in December. 

So the Stargate announcement now looks like another example of U.S. tech leaders hurrying to curry favor with Trump, in this case by offering him a chance at some unearned political points. That’s not to say the demand for new data centers isn’t real. “I generally think that there needs to be (and will be) a massive infrastructure buildout to really ‘enable’ AI to reach its potential,” says Anyscale cofounder Robert Nishihara in an email. “This has been true of every major technology wave—internet, cloud, mobile—we’re just at the start.” 

And yet the hardware may play a unique role in this new tech wave compared to previous ones. The generative AI boom owes a lot to the availability of powerful GPUs. The game-changing intelligence gains that led to ChatGPT happened mostly because researchers discovered ways of applying more computing power to large transformer models. With enough GPUs, something remarkable happens (researchers don’t know exactly what) deep in the layers of the neural network that results in an alarmingly nuanced understanding of the relationships among words, code, or images. Even more surprising was the discovery that adding more and more GPUs leads to better and better results. Altman is betting OpenAI can ride this scaling law all the way to artificial general intelligence, then superintelligence. “If that holds,” Nagaraj says, “it effectively means that whoever controls the infrastructure will control a lot of the market.” 

Chinese DeepSeek model makes strong case for state-of-the-art AI

A small Chinese research lab called DeepSeek has sent shockwaves through the AI world by releasing a new reasoning model called DeepSeek-R1 that rivals OpenAI’s o1 model in mathematics, coding, and general knowledge, according to benchmark tests. The model can break down problems into chunks, and scrutinize and verify its own results. DeepSeek reportedly costs 90-95% less to run than OpenAI’s o1. 

Remarkably, the Chinese company open-sourced the models, meaning developers can freely access and modify them for their own uses. In other words, DeepSeek is doing what OpenAI set out to do when it started in 2015: developing “open” AI models that can be used to benefit all of humanity. OpenAI has since locked down the details of its most capable models, and charged developers fees to use them. So why would DeepSeek give away the value it created with DeepSeek-R1? 

“It helps DeepSeek attract top talent by showcasing their technical capabilities and commitment to collaborative development,” says David Bader, a professor at the New Jersey Institute of Technology. “The strategy also fosters a developer ecosystem around their technology, potentially accelerating innovation and adoption.” In the end it could position DeepSeek for broader influence in both domestic and international AI development, Bader says. DeepSeek has emerged as a leading force in China’s language model development, surpassing established tech giants like Baidu, Tencent, Alibaba, and ByteDance, Bader explains. DeepSeek has also set off a price war among China’s AI providers.

DeepSeek is similar to o1 in some ways. For example the model does lots of computation at “test time” after it’s been prompted by a user to begin work on a problem. It can feel its way through various approaches to the problem, following promising pathways while steering away from nonproductive ones. But this doesn’t mean that DeepSeek “copied” OpenAI’s work on o1. “[T]he core knowledge for building these models is taught in universities and discussed openly at conferences,” Bader says. “When developments like OpenAI’s o1 model emerge, it’s natural for other implementations like DeepSeek’s R1 to follow, as researchers worldwide build upon shared technical foundations.”

Accenture: More AI budget should go to training humans

Accenture, which helps enterprises adopt AI, is out with results from its latest round of survey work on executive and worker views on AI deployment in 2025. The firm surveyed 3,450 C-suite leaders and 3,000 employees across 22 industries in 20 countries in the last quarter of 2024.

A few highlights from the survey:

• “2024 was the year of gen AI—but after 12 months of rapid adoption, just 50% of C-suite leaders claim to be fully prepared for technological disruption and only 36% say they have scaled gen AI solutions. A mere 13% report seeing significant enterprise-level value.”
• “The top three reasons C-suite leaders cite are capitalizing on advances in the technology (28%); maintaining business competitiveness (20%); and increasing confidence in managing associated business risk (20%).”
• “Just 12% [of executives] consider improving skills as the main [reason to invest]. This is corroborated by Accenture’s own experience working with companies globally: the company found that three times more gen AI budget is spent on technology than on people.”

More AI coverage from Fast Company: 

• States are turning their public benefits systems over to AI. The results have often led to ‘immense suffering’
• ‘The Brutalist’ used AI voice editing. Is that such a crime?
• Why Trump dashed Biden’s AI executive order, and why it matters
• Samsung’s Galaxy S25 phones are—you guessed it—all-in on AI

Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.

https://www.fastcompany.com/91265565/the-oval-office-stargate-project-reveal-was-just-more-tech-industry-genuflecting?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss