DeepSeek's Disruption and Impacts on the Tech Industry
by Apollo Colligan
President, Business and Economics Club
Artificial intelligence has been advancing at a breakneck pace and has completely revolutionized almost every aspect of modern life, with U.S. firms leading the charge—until now. DeepSeek, a once-obscure Chinese AI firm, emerged onto the scene in early 2025 with a model that rivals the best in the industry, at apparently a fraction of the cost. The impacts on the AI industry, and the way we interact and understand AI, has completely changed. Many argue that this disruption will be incredible for AI. Others warn that DeepSeek is nothing but trouble. However, I believe the truth lies somewhere in between.
DeepSeek began development in May 2023 when it was founded by Liang Wenfeng in Hangzhou, China. By May, the U.S. AI market was worth a combined $42 billion, with the AI index (combined U.S. AI stocks) rising 29.29% YTD (year to date) from January. The company quickly began releasing large language models (LLM) when they released DeepSeek Coder, an open-sourced model specifically designed for coding-related tasks in November 2023. By November, the U.S. AI market grew to a combined worth of $123.07 billion with the AI index rising 75.92% YTD. By December, DeepSeek launched its first general purpose LLM. The United States continued to dominate, reaching all-time highs in evaluation, as bolstered by increased public awareness and new Department of Defense contracts. The AI index concluded 2023 with 78.09% increase YTD.
Graph of S&P 500 price over time, by macrotrends.net
During 2024, as U.S. AI and tech stock skyrocketed, Deepseek continued to further its development without many public releases. Its advancement was hampered by advanced chip unavailability and proprietary U.S. AI software. It is around here when many suspect DeepSeek and China as a whole of smuggling in NVIDIA chips through Singapore and stealing OpenAI’s IP. Despite these developments, the AI market in the United States continued to surge. 2024 was characterized by increased regulation of the AI market in the U.S. and its integration into many large companies and organizations, as seen in its adoption by NATO, the Mayo Clinic, Cisco, Department of Homeland Security, and various social media companies. By May 2024, DeepSeek released DeepSeek-V2, which focused on lower training costs. Although its capabilities remained inferior to those of its U.S. competitors, its lauded low development price was very competitive compared to the U.S. AI models.
On January 20, 2025, DeepSeek released DeepSeek-R1, whose performance matched or exceeded ChatGPT-4o, its leading competitor and ChatGPT’s most advanced model. The claimed low development price of only $5 million sent the U.S. AI and chip market into freefall. The tech-heavy NASDAQ dropped 3% and NVIDIA, the chief supplier of chips used in AI training, fell an immense 17%, losing $600 billion in market capitalization in just a single day. DeepSeek’s capabilities included creative cognition on par with ChatGPT and the ability to more easily integrate into coding services. DeepSeek claimed to have only been trained for $5 million but the number may be closer to $50 billion — on par with U.S. AI models.
Following the stock-shock of January 20th and the swift adoption of DeepSeek by American consumers, multiple investigations were launched by the U.S. government and various American companies on whether DeepSeek used smuggled NVIDIA chips and OpenAI’s IP. The White House “AI czar” David Sacks stated that DeepSeek was trained using OpenAI: “[There is] substantial evidence that what DeepSeek did here is they distilled knowledge out of OpenAI models, and I don't think OpenAI is very happy about this." (Data Center Dynamics, 2025). The FBI also suspects DeepSeek of using NVIDIA chips smuggled in via Singapore to China. NVIDIA sales to Singapore increased from $7.6 billion in 2023 to over $17 billion in 2024, accounting for 22% of NVIDIA’s worldwide revenues. With the information that DeepSeek was not cheap nor trained on non-NVIDIA chips, the tech sphere is recovering, with NVIDIA advancing steadily upwards since the dip.
Many leading tech companies, including Microsoft, have stated that DeepSeek’s disruption will be good for the tech sphere in the long term. They believe that the introduction of a competitive model at cheaper prices demonstrates the viability and compatibility of modern LLM technologies. However, there is little chance that DeepSeek will be adopted by large-scale buyers due to security vulnerabilities and the fear of Chinese surveillance. According to HackREAD, “DeepSeek-R1 LLM fails 58% of jailbreak attacks in Qualys security analysis.” This means that an intelligent user could extract sensitive information from any backside engineering, demonstrating a huge lack in security. Additionally, the lower relative cost of DeepSeek is having a revitalization effect on the tech industry. Microsoft CEO Satya Nadella stated in a recent investor report that DeepSeek will increase the ubiquity of AI and thus help grow the AI market as a whole.
The possible innovation of DeepSeek, in conclusion, will be good for the AI industry as a whole. Many believe that it has “lit a fire” under the U.S. AI market to maintain its global domination of the tech industry; others believe that this could be the start of a new space race. Others think that this is the pinnacle of AI tech as a whole, and the only further innovations that can be made are to make the tech cheaper. The only thing we can do is wait and see.