Deepseek: A journey from Hedge Funds to AI

Introduction:

In this busy and bustling day to day life of ours managing our Finances has become a great deal of work for us. From investing in Mutual Funds to Equity and other instruments we have to go through a lot of information around us, which sometimes might turn out to be a rock on a glass. Industries have enhanced; new industries have begun and likewise the markets have also become a critical battlefield in itself, with waves of public money heading ups and downs. With the advancement of modern technology and especially in the age of AI a lot of things have changed with the introduction of newer models in finance. From OpenAI’s Chat Gpt to Aravind Srinivas’s Perplexity, researching and doing analytics on Finance has become an easier task. But though the task has become easier, we still couldn’t rely only on those. While there are many AI driven companies that have emerged recently, one such company that is now a grossing topic between the netizens is “DeepSeek”. Founded by a hedge fund “High-Flyer” it has been a powerful or equivalent AI than most of the already existing generative AI models. I bet you have heard this name since the start of the year, but do you know that the potential it has to outperform the rest of the companies is likewise commendable. So let us dive deep into the epoch of DeepSeek and how its presence has been a blessing or a facade in disguise. 

Journey from the East:

Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd doing Business as DeepSeek is a Chinese Artificial Intelligence Model which was founded by the hedge fund High-Flyer group as a part of their research project. High-Flyer was co-founded by Liang Wengfeng, who now serves as the CEO for both the companies. As a generative artificial intelligence model it is also responsible for developing large language models. It all started on 14th of April 2023 when High-Flyer announced its plan of developing an artificial general intelligence lab dedicated to research regarding AI tools and models which was separate from High-Flyers financial business. It was incorporated on 16th May 2023 with High-Flyer backing as an investor and being the parent company, the lab became its own and came to be known as “DeepSeek”. 

So what’s the thing about DeepSeek that differs it from the rest of its competitors in the market? Well before we move to that topic, let us see the models that DeepSeek has developed over the period of time.

One of its first models was the DeepSeek Coder whose architecture was similar to the Llama Model. But the model that has been on the news and the feed of Netizens is the R1 model. So what’s the catch regarding the R1 Model? In simple terms we can say it’s one of the cheapest and most competent models in the market and this has led to one of the most surprising bearish turnout in the US market. The speed at which Chinese AI company Deepseek has shaken the technology sector, the markets and the bullish behaviour of the Americans is nothing short of stunning. Marc Andressen, a renowned Venture capitalist also said on X that, “DeepSeek-R1 is AI’s Sputnik moment” comparing it to the satellite that kicked off the space race. And recently it became one of the most downloaded apps on Apple’s US app store. The new AI chatbot has triggered massive selling of some major tech stocks which were already in free fall as America’s leadership was questioned in this sector. 

An Earthquake that jolted Wall Street for a day:

As we have seen AI evolve over time and how it has developed to become more and more reliable; tech companies who are working on making these kinds of technology are spending billions and billions of money accounting up to $320 Billion for the year 2025, just to build large Data centres powered by NVIDIA’s GPUs and then outperform in the market. Amazon alone has spent more than $100 Billion for AI, and they considered it as “once in a time of business opportunity.” Microsoft has invested $80 Billion for AI data centres, Alphabet has invested $75 Billion for AI infrastructure and Meta has invested around $60-65 Billion for AI growth. Whereas, Apple is investing in AI through cloud partnerships and Tesla has allocated $5 Billion for AI development and NVIDIA here is acting as a supplier of high-performing GPUs,and not as a buyer. Now if we compare DeepSeek and any one of the tech giants, we could notice that DeepSeek has not even spent much amount and yet it is outperforming most and being equal to some of the AI interfaces. So what is DeepSeek’s secret for its frugal success? DeepSeek followed a method called the “the method of mixture” where rather than following the traditional method of learning everything through one giant neural network, Deepseek instead split it into different small networks. One for poetry, one for code and so on. Each of them acting as an “expert” in their domain and a “generalist” acted as a bridge, connecting them together. Now one more question may arise that if this was so simple, why did the big tech giants like OpenAI, Gemini by Google follow this method even after having people with brains and vast monetary resources? That is because building AI models is expensive and trying new techniques may burn millions and billions of money with no guarantee of success in return. But DeepSeek took the gamble and found their fortune. Another reason for them to outshine is the cost-effective chip that they produce and use. To people’s surprise the latest model R1 was only built with $5 Million in comparison to $100 Million that was spent by OpenAI to build ChatGpt. People never thought that AI models could be trained and utilized on normal and cheap chips and after seeing DeepSeek do it with ease there has been a conundrum regarding NVIDIA’s costly chips. In late January 2025, Nvidia and Broadcom saw a sharp decline in their stock prices following the emergence of DeepSeek. The development of DeepSeek to produce models rivaling top U.S. AI systems at a fraction of the cost, without relying on advanced foreign processors triggered fears of a shift in AI infrastructure investments, causing Nvidia’s shares to drop by 17% and Broadcom’s by 17.4% on January 27. However, the market reaction was partially reversed the next day, with Nvidia rebounding by 8.9% and Broadcom by 0.9%, as analysts suggested the sell-off was likely an overreaction, given the continued strong demand for AI-related hardware. While this was just a glimpse of how the market turned out to be due to the announcement of just one model. But this can also let us predict what we may call the coming of the “AI Bubble”.

Recent Backlash and Controversy:

While DeepSeek is performing at its best capacity with and with the recent developments it seems to outperform literally the best in the market too. But lately DeepSeek has been facing backlash due to privacy concerns. Italy has banned its services after DeepSeek failed to address concerns regarding its privacy policy. It also increased Geopolitical tensions between China and Taiwan as Taiwan banned the use of DeepSeek services in school, government agencies and state-owned enterprises due to the risk of data leakage and transmission from one state to another. Australia banned the usage of DeepSeek on all federal devices due to the risk of security and similarly in the USA, most of the federal institutions have restricted DeepSeek. South Korea’s Minister of Trade and India’s Finance Minister have asked government officials to restrict the use of AI and most importantly DeepSeek in a government setting, citing risk regarding government documents and data. 

The Silk Road of AI:

The story of DeepSeek is no less than a movie scene where the protagonist with no resources at all rises up to power against all the antagonists in the movie scene. Geopolitically if we draw parallels in the eastern region of Asia, there has been a great development in the spectrum of AI. While the tech dominant Silicon Valley of the west has always been known for its genius brains and rapid technological developments, the developers of the east have paved a way to prove their technological intellectuality. Not only the development of DeekSeek, but there are many AI Software that has been developed and is in the development stage here in the east of Asia. The Asian markets are flooding with tools and invention of AI powered softwares, and we might see this development as a counter attack to the western dominance .

What I think about It:

In the end what I think about DeepSeek is that it is one of the finest creations in its field any group of humans have done till date. AI has a lot of potential in Finance and we may see the rise of a lot of Fin-Tech companies this year and in the upcoming years. Usage of AI in algorithmic trading, robo-advisors and overall analysing and predicting market performance will be a potential skillset. Even though DeepSeek has shaken the American firms, this AI war won’t stop here. We may see more and more developed and excruciating AI models in the future that could then do tasks that we can’t generally think about now. And there might be development of AI in Finance that will change investment, but we still can’t say for the good or the bad.

Kabir Kashyap

Member

Session 5

Session 5- Unraveling Equity Derivatives: Insights from the Fifth Development Session The Fifth Development Session…

Read More

Leave a Comment

Your email address will not be published. Required fields are marked *