
- DeepSeek-R1 is making waves as a faster, smarter, and cheaper alternative to ChatGPT and other LLMs »
- DeepSeek’s training cost less than $6 million, which is 6% of OpenAI’s ChatGPT-4 investment »
- Many social media reactions, from both consumers and industry leaders, are generally positive »
- There is concern about cybersecurity risks due to its open-source nature and ties to China, especially during the Tech Cold War »
In a single day, one small startup went from no recognition to suddenly dominating the stock market, Google Trends, app charts — and seemingly everyone’s news feed.
That startup is called DeepSeek, a Chinese AI company that just released its third iteration of its large language model (LLM), which is said to rival OpenAI’s ChatGPT.
Its arrival is making big waves, notably because DeepSeek recently announced how much it actually cost to train its new model, DeepSeek-R1. The total was less than $6 million — an eyebrow-raising difference when compared to OpenAI, which invested more than $100 million into its most recent model, GPT-4.

President Trump quickly weighed in, calling the launch “a positive development,” and credited DeepSeek for being the first to make a faster AI at a lower cost. But, he added, it should be considered a “wake-up call” for the U.S.
There’s reason for the congratulations. According to some findings, DeepSeek’s performance hits benchmarks similar to those of OpenAI’s latest model.
In some areas, though, DeepSeek surpasses ChatGPT, particularly in mathematics and coding. According to Tom’s Guide, DeepSeek also had faster and more accurate responses in complex problem-solving tasks.
OpenAI’s CEO, Sam Altman, also remains optimistic, praising R1 as an “impressive model, particularly around what they’re able to deliver for the price,” but not without adding that OpenAI “will obviously deliver much better models.”
Marc Andreessen, a Silicon Valley venture capitalist whose Netscape Navigator helped trigger the internet era some 25 years ago, posted on X that “DeepSeek-R1 is AI’s Sputnik moment.”
But not everybody is as thrilled.
Nvidia takes a hit
The release of DeepSeek-R1 comes just after former President Biden implemented a rule restricting the sale of AI chips to certain countries, including China. Among those affected is the world’s largest chip manufacturer, Nvidia.
But DeepSeek used Nvidia chips — just not the high-end ones that most sophisticated LLMs, like ChatGPT, use. Instead, it opted for Nvidia A100 and H800 chips, which are less powerful than the newer H100 models.
Alexandr Wang, CEO of Scale AI, told CNBC that he believed DeepSeek had access to 50,000 H100 Nvidia chips, which would be in breach of U.S. export rules. Whether that’s true remains to be seen, but it would be interesting to see the kinds of legal implications that could play out for the Chinese company if so.

For Nvidia, more bad news was on its way. Aftershock rolled in on Monday, January 27, when its stock dropped by more than 17%, costing the company $589 billion in market capitalization. It was the biggest single-day loss of value for any public company in history.
As it becomes clear it’s possible to build a sophisticated model for less money and with less powerful chips, there’s obviously a question of whether there’s a future for Nvidia’s high-end GPUs.
The good news
Kevin Baragona, founder of DeepAI.org, said this is an opportunity to “challenge the idea that AI breakthroughs are reserved for the wealthiest companies.”
“DeepSeek is making waves in AI, proving that you don’t need Silicon Valley budgets or cutting-edge hardware to build something impressive,” Baragona said.
There’s some uncertainty about whether DeepSeek’s $6 million investment — which represents just 6% of OpenAI’s total investment in its own chatbot — is totally accurate.
“DeepSeek employs a lot of people and is relatively well-funded, proving that money isn’t everything. There’s also the possibility they receive more government funding than is publicly disclosed, which could distort the narrative,” Baragona said.
There definitely could be more layers to DeepSeek’s surprisingly small $6 million investment that may come out in the wash. And given the Nvidia drop, some people may be jumping to conclusions a little too quickly.
Microsoft’s CEO, Satya Nadella, compared DeepSeek’s prowess to the Jevons paradox in a LinkedIn post, reading: “As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can’t get enough of.”
Jevons paradox centers around the idea that technology, even if it becomes cheaper or more efficient, can actually lead to more widespread use.
The shift opens up more opportunities for AI to become accessible to a broader range of users. It could force “bigger companies to reconsider their approach — AI development may no longer be just about who spends the most,” Baragona said.
Reactions to DeepSeek have been, for the most part, overwhelmingly positive, with many users praising the LLM for being nearly free.
DeepSeek costs between $0.14 and $0.55 per 1 million tokens, which could equate to 100,000 queries for less than $1. A similar plan on ChatGPT can cost up to $200 per month.
Others are happy to see that DeepSeek is open-source, signaling to everyone everywhere that the company is open to collaboration on its future models. DeepSeek’s R1 and V3 models are available on GitHub.
In the face of the Tech Cold War
All of this comes at a time when the geopolitical landscape is growing tense.
Whether you call it the New Cold War or the Tech Cold War, competition is at an all-time high between the U.S. and China as both countries attempt to make their way to the top of the metaphorical AI mountaintop.
While President Trump expressed optimism, concerns about data privacy and China are amplified — as many of us have seen in the ongoing U.S.-ByteDance controversy.
The issue centers on fears that TikTok, owned by ByteDance, could be sharing sensitive user data with the Chinese government. The fact DeepSeek is Chinese-owned (and open-source) could pose a similar cybersecurity risk.
Former President Biden emphasized this in his executive order “Strengthening and Promoting Innovation in the Nation’s Cybersecurity” just before he left office. It pushed for stronger security around open-source software and recommended government agencies work together on best practices for managing open-source projects.
Just a few days later, President Trump rolled back the order. But that’s not to say he isn’t helping pioneer an AI revolution.

Perhaps he thought Biden’s action plans were too hypervigilant because, during his entire campaign, President Trump made a point of supporting the American AI industry. After gaining office, he immediately signed an executive order calling for the creation of an AI Action Plan that is aimed to “solidify [the U.S.’s] position as the leader in AI.”
There’s no doubt that DeepSeek just threw a curveball into America’s plans to race to the top. But it is likely just that: one good pitch in a very long game where the other player has the talent to swing back.
“Whether DeepSeek’s success is purely a result of ingenuity, or if other factors are at play, remains an open question,” Baragona said.