Chinese AI startup DeepSeek’s artificial intelligence (AI) models seem to perform on par with OpenAI and Google’s leading models. Still, with far fewer resources—the result was a steep 15.3% dive for Nvidia stock as Wall Street reckons with what this could mean going forward.
Are Investors Overpaying for AI?
The concern is that investors could be overpaying for the emergent technology since DeepSeek emulated the successes of major American tech companies with less funding and no access to Nvidia’s highest-grade chips. The Chinese startup has released two model families: the V3 AI chat and R1 reasoning models.
DeepSeek claims that R1 cost them $5.58 million to make and used 2,048 of Nvidia’s H800 chips for training.
On the other hand, Anthropic CEO Dario Amodei stated that AI models cost $100 million to train but could cost as much as $100 billion and would use thousands of high-grade chips.
An Impact on Upcoming Plans in AI
The cost disparity comes at a bad time for OpenAI, SoftBank, Oracle, and MGX. The AI companies recently announced plans to spend between $100 billion and $500 billion on building AI infrastructure for project Stargate, which the United States government would be involved in. The cost estimates are in question, resulting in stock drops from Google, Microsoft, and Amazon.
“Artificial intelligence has reached a critical inflection point,” senior manager at SAP Gokul Naidu told PYMNTS. “The industry stands at a crossroads where escalating costs, environmental concerns, and innovation appear intertwined, threatening to stifle accessibility and adoption. Enter DeepSeek-R1, the model that’s turning heads in Silicon Valley and beyond for proving that high performance and affordability aren’t mutually exclusive.”
DeepSeek: Copycat or Breakthrough?
Though DeepSeek has become the most downloaded app on the Apple App Store, OpenAI CEO Sam Altman responded in an X post when DeepSeek’s V3 model was first released a month ago. “It is (relatively) easy to copy something that you know works,” he wrote. “It is extremely hard to do something new, risky, and difficult when you don’t know if it will work.”
Other leaders have a different opinion. “DeepSeek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen,” Marc Andreessen of the Andreessen Horowitz VC firm posted on X, “and as open source, a profound gift to the world.”
Meta Chief AI Scientist Yann LeCun echoed the sentiment in a LinkedIn post that addressed concerns that China would overtake the United States in AI. “The correct reading is: open source models are surpassing proprietary ones… they came up with new ideas and built them on top of other people’s work. Because their work is published and open source, everyone can profit from it.”
Finding Success Despite Severe Limitations
Another consideration is how DeepSeek circumnavigated former President Joe Biden’s export controls on advanced AI chips, which outline access levels for more than 120 countries. DeepSeek could not access the latest Nvidia GPUs, so it had to rely on older models. Despite this, the model has outperformed other expensive models in some tests.
“This feat speaks volumes about the ingenuity behind DeepSeek’s approach,” Naidu continued. “It’s not just about throwing money at the problem; it’s about finding smarter, leaner ways to train and deploy AI systems.”
“DeepSeek is more than a model—it’s a wake-up call for the entire AI industry,” Naidu concluded. “It challenges entrenched assumptions about the cost of innovation and offers a path forward where cutting-edge technology is both that are affordable and sustainable. As we move deeper into 2025, the conversation around AI is no longer just about power—it’s about power at the right price.”
