NEW ORLEANS – Just when the dust was beginning to settle on the top AI companies such as OpenAI, Google, Microsoft, and Nvidia, a new kid arrived on the block and it took the AI industry by storm.
DeepSeek, a Chinese artificial intelligence company based in Hangzhou, Zhejiang, rapidly positioned itself as a significant player in the global AI landscape when it was released on Jan. 27. Established in 2023 by Liang Wenfeng, co-founder of the hedge fund High-Flyer, DeepSeek has developed open-source large language models (LLMs) that rival those of leading Western firms.
“DeepSeek’s entrance to the market is like the equivalent of an 11-year-old making it into the NBA,” said Henry Hays, CEO and founder of DisruptREADY. “Impossible!”
DisruptREADY is a strategic advisory firm that helps executive teams keep up with rapidly evolving technologies such as blockchain, crypto, and Web3.
Envelope-Pushing Efficiencies
DeepSeek’s latest AI model, DeepSeek-R1, performs tasks at a level comparable to OpenAI’s ChatGPT, but achieved this with a much lower development cost, and requires only a tenth of the computing power typically needed for comparable LLMs.
“Open AI spent four years training ChatGPT 3, spending over $100 million,” Hays said. “DeepSeek spent around $5.5 million with just 200 employees. To put the pace of change into perspective, Nextflix, one of the original gangsters in the streaming service arena, took 3.5 years to get to 1 million paid subscriptions. It took ChatGPT just five days. That was two years ago. Now with DeepSeek we’re dealing in a space and speed we’ve never seen before.”
On Jan. 27, DeepSeek-R1 surpassed ChatGPT as the most-downloaded free app on the iOS App Store in the United States. This rapid increase led to a reported cyberattack, prompting DeepSeek to limit new user registrations.
DeepSeek’s cost-effective achievement raised concerns among U.S. tech companies and investors. According to Business Insider, following the debut of DeepSeek’s AI model, shares of Nvidia plunged nearly 18%, leading to a broader decline in tech stocks. The Nasdaq 100 fell about 3%, and the S&P 500 decreased by nearly two percent.
By the afternoon, though, Nvidia’s stock had rebounded by over 8%, closing at $128.86. This recovery was mirrored across the tech sector, with indices like the S&P 500 and Nasdaq Composite gaining approximately 1% and 2%, respectively.
Analysts suggest that the initial sell-off was driven by fears of increased competition and potential reductions in AI-related spending. However, the subsequent rebound indicates investor confidence in the sustained demand for existing AI infrastructure and services. The market’s quick recovery underscores the belief that the emergence of more efficient AI models could lead to cost reductions and wider adoption, potentially benefiting the industry as a whole in the long term.
Open to the Public
Hays, who is also a part-time Adjunct Professor at the Louisiana State University Honors College, teaches students how to build LLMs in a program sponsored by companies like Shell, BASF, Google, Our Lady of the Lake, and Entergy. He says that what DeepSeek is doing is unique in more ways than one.
DeepSeek decided to open-source its advanced AI DeepSeek-R1 and DeepSeek-V3 models on GitHub which is free and open to anyone. “GitHub is ground zero for coders,” says Hays. “DeepSeek’s introducing their AI code on GitHub is a technological minor miracle.”
DeepSeek’s decision to share its code on GitHub has sparked a mix of commendation and apprehension among leading AI companies, highlighting the complex dynamics between open-source collaboration and competitive advantage.
AI Agents
The future of AI agents is rapidly approaching a point where they will seamlessly integrate into everyday life, understanding and predicting consumer behavior with incredible accuracy.
“Just like J.A.R.V.I.S. used by the fictional Iron Man character, Tony Stark, in the Marvel Universe, we will have AI Agents in real life,” said Hays. “It’s closer to reality than most people think.”
These agents will be able to understand detailed consumer preferences, such as size, color preferences, and purchasing habits, while anticipating needs like delivery services. As AI agents become mainstream, the demand for alternative energy sources to power the vast data processing will intensify.
Increased Energy Use Means Looking at the Nuclear Option
The significant energy demands of AI systems present a challenge to their scalability and sustainability.
“The soft underbelly of the AI ecosystem is the amount of energy it takes to use it,” said Hays. “If you google New Orleans pizza it takes a certain amount of energy to generate a result, but if you use AI the energy used is around thirty times more. DeepSeek won’t change the fact that we need data centers to securely store and process AI data, but the company has demonstrated that the use of AI can be much more efficient.”
The adoption of nuclear power is gaining renewed attention as a viable option to meet the massive and increasing need for clean energy to fuel AI. As a sign of this trend, Amazon purchased Three Mile Island for over $500 million in Oct. last year to explore the development of a small modular nuclear reactor (SNR).
“Modular SNRs are now the size of an RV,” said Hays. “Soon big employers will have their own SNRs on site to run AI Agents and they won’t be beholden to any specific utility.”
If businesses don’t jump on the AI wagon now, they may be left behind as commerce rapidly evolves to integrate AI and AI Agents. DeepSeek’s achievements highlight this evolving landscape and the increasing prominence of firms that can move quickly and more efficiently in this critical sector.
