DeepSeek, a Chinese company, released a new LLM that literally disrupted the AI space and crashed the global financial market. Well, we will talk about this in a minute.
Ever since OpenAI announced its arrival by releasing ChatGPT in November 2022, there has been a disruption across every industry as we know it. OpenAI made ChatGPT a household name, and terms such as “AI,” “gen AI,” and “generative AI” became mainstream.
Over the years, the LLMs powering ChatGPT and its competitors from Google, Anthropic, and Meta have improved significantly. Only Anthropic’s Claude has come close to challenging ChatGPT in some benchmarks, while Google’s Gemini has recently outperformed both ChatGPT and Claude with their latest models.
While OpenAI has been releasing state-of-the-art updates to maintain its dominance, a new competitor from China, DeepSeek, has entered the AI landscape with the sole intention of challenging OpenAI like never before.
DeepSeek — the OpenAI Killer (literally)
On December 5, 2024, OpenAI released its flagship model, o1. This model was trained using a large-scale reinforcement learning algorithm that enables it to answer queries effectively by employing the Chain of Thought (CoT) technique.
CoT allows AI language models to break down complex problems into smaller, manageable steps and provide clear explanations for each step. In simpler terms, when prompted with a query or task, the model takes a moment to “think” and then responds appropriately. Most of the time, its responses are accurate or close to accurate.
OpenAI charges $20/month for its standard subscription and $200/month for its Pro subscription.
On January 20, 2025, DeepSeek launched R1, its first reasoning model, which parallels OpenAI’s o1.
DeepSeek was founded by Liang Wenfeng, a former hedge fund manager who previously worked at High-Flyer Quantitative Investment Management.
The startup’s latest model, DeepSeek-R1, is priced at approximately 2% of the cost of its closest competitor. It was trained using approximately $6 million worth of Nvidia H100 GPUs and is faster, cheaper, and as accurate as OpenAI’s o1. Just for comparison, the startup has less than 200 people working for them, whereas OpenAI has 4500 employees.
While the model is free for general users, the API pricing for R1 is $0.55 per 1 million input tokens and $2.19 per 1 million output tokens, compared to o1, which costs $15 per 1 million input tokens and $60 per 1 million output tokens.
This makes R1 significantly more affordable for developers and businesses. This development has disrupted a significant revenue stream for OpenAI.
#1 on App Store
The DeepSeek chat assistant has reached the number one spot on the App Store, surpassing even ChatGPT.
DeepSeek is having the ChatGPT moment where it is surpassing every other app in the app store. This is working because it is free and open-source for the people.
The most remarkable aspect of the situation is that the former US President had issued several executive orders aimed at preventing companies like NVIDIA from selling their GPUs to China. The goal was to hinder China’s progress in the AI race by limiting their ability to train powerful models.
However, the outcome was quite different: it pushed companies like DeepSeek to excel at developing compute-efficient large language models. You have to appreciate that.
Data Privacy & Censorship concerns
DeepSeek is a Chinese company based in China. Anything you ask or share via the app or website will be shared to the servers in China. This raises significant data privacy concerns. Read the terms of service, and you will know. This can be mitigated by running the downloaded models on your local machine using Ollama or LM Studio.
Another issue with this chat assistant is that it censors certain information due to the guardrails in place, which is a standard procedure. Although this can be bypassed.
Reactions across The World
A few years ago, Sam Altman was asked whether a startup with a smaller budget, such as $10 million, could make a significant impact in AI. Altman replied, “It’s totally hopeless to compete with us on training foundation models.”
Call it confidence or call it arrogance, but this was the reality at the time.
Who knew that a startup less than 2 years old would create an AI product or ChatGPT competitor that would outperform OpenAI’s counterpart in some aspects?
Marc Andreessen, a tech community icon whose Netscape Navigator played a pivotal role in the internet revolution, is drawing a historical comparison to 1957, when the Soviet Union shocked the U.S. by launching the first satellite into orbit.
“DeepSeek-R1 is AI’s Sputnik moment,” Andreessen posted on X.
Aravind Srinivas, CEO of Perplexity, has also integrated DeepSeek models into Perplexity, an AI-powered search engine and chatbot. He posted, “I’m obsessed about DeepSeek R1 and Search. Can’t stop using it. Web, internal files, third-party apps, tools, …everything.”
Altman later posted that he was impressed with R1, particularly with what it offers for the price.
Altman also promised that OpenAI will deliver better models, and he is excited to have a new competitor.
Industry professionals have come out in full support of DeepSeek after it faced criticism for being a Chinese startup.
In fact, people have been praising the fact that a startup less than two years old, with its open-source model, can challenge an AI tech giant like OpenAI.
Nvidia’s Biggest-ever $600 billion Market Crash
Unlike AI models in the US that depend on costly chips and enormous energy resources for training, DeepSeek delivers competitive performance with far less power and more budget-friendly hardware.
This is precisely the type of disruption that can disrupt the tech landscape and unsettle markets. And it did.
Nvidia Corp. has suffered the largest single-day market value loss in U.S. stock market history, wiping out $600 billion amid investor concerns over the rise of DeepSeek.
On January 27, 2025, Nvidia’s shares tumbled by as much as 17%, exceeding its previous record drop of $279 billion in September of the previous year. The broader tech sector also took a massive hit, losing nearly $1 trillion ($969 billion) in market value.
I feel DeepSeek is a great product, and they deserve all the praise they’re getting. However, I still believe OpenAI’s o1 is a better reasoning model. The problem is that you can’t use it for free, and even if you do, you’ll face rate limits. With DeepSeek’s R1, that’s not an issue.
Many people never paid for OpenAI’s services, but when they got a taste of a reasoning model like R1, they realized it was a game-changer. No wonder the app was downloaded at lightning speed. We’ve seen a similar scenario before. In November 2022, people adored ChatGPT because, for the first time, they felt empowered to create and develop using AI.
ChatGPT started as a free tool, and while it remains free, its best features are now locked behind a subscription. The issue? That subscription is too expensive for many users.
And that’s exactly where DeepSeek capitalized. The tech landscape is evolving every day, every hour. You never know what comes next. It’s an exciting time to be alive.
If this article provided you with value, please support me by buying me a coffee—only if you can afford it. Thank you!
