The artificial intelligence arms race just got more competitive. Chinese AI startup DeepSeek has caused quite a stir in both tech and financial circles this week, rolling out its R1 reasoning model and quickly climbing to the top of Apple’s App Store charts in the US. The app’s sudden popularity bumped OpenAI’s ChatGPT from its usual spot, sending ripples through the market—ripples that turned into a tsunami for Nvidia, whose stock dropped 17 percent in a single day. That's nearly $600 billion in market value erased overnight, the largest single-day drop in US corporate history, according to CNBC.
DeepSeek's rise comes at a curious time. Just days ago, Donald Trump announced the Stargate Project, a $500 billion initiative involving major tech companies like OpenAI, Oracle, and SoftBank to bolster US AI infrastructure over the next four years.
"What we want to do is we want to keep it in this country," Trump said. "China is a competitor, others are competitors. We want to be in this country, and we're making it available. I'm gonna help a lot through emergency declarations, because we have an emergency, we have to get this stuff built. So they have to produce a lot of electricity. And we'll make it possible for them to get this production done easily, at their own plants if they want."
The project promises thousands of new jobs, cutting-edge innovation, and, in Trump’s words, a way to “beat China” in the AI race. But DeepSeek’s emergence threatens to undercut these efforts. If its claims of building high-performing AI models at a fraction of the cost with fewer GPUs are true, it could render much of the Stargate strategy moot before it even gets off the ground.
Founded in 2023 by Liang Wenfeng, DeepSeek has been making noise about its ability to do more with less. Its models, including the new R1 and the just-released Janus-Pro multimodal AI, are open source, cost-efficient, and, apparently, quite capable, even though they reportedly require significantly fewer resources and cost less to train compared to the world’s leading models. Janus-Pro, for instance, claims to outperform heavyweights like Stable Diffusion and DALL-E 3 in image generation benchmarks and it's new R1 reasoning model is designed for solving complex problems and is said to perform as well as OpenAI’s o1 on certain benchmarks, according to The Verge.
Adding to the drama, DeepSeek has begun restricting signups for its app, citing “large-scale malicious attacks.” Existing users can breathe easy for now but would-be downloaders might need to wait in line. Meanwhile, the incident report reads more like a teaser than an apology, inviting users to “try again” later. Whether the attacks are real or just part of a savvy PR strategy, they’ve only added to the buzz.
Even OpenAI’s CEO, Sam Altman, weighed in. Calling the R1 model “impressive,” Altman struck a diplomatic tone but made it clear his company won’t be standing still. “We will obviously deliver much better models,” he said, before pivoting to a classic conversation about AGI, as one does when their app has been dethroned.
DeepSeek’s rise comes at a time of heightened tension in the AI world, especially between the US and China. As for Trump’s Stargate Project, it’s starting to look less like a bold new frontier and more like a very expensive game of catch-up.
