DeepSeek Statistics 2025 – AI That Wiped Out $1 Trillion

DeepSeek R1 steals the show by beating ChatGPT in every benchmark. This led to a $1 Trillion loss in the stock market in a single day.

Editor’s Top Picks:

• DeepSeek’s Popular AI App Is Explicitly Sending US Data to China (Wired.com)
• Deepseek’s Founder Liang Wenfeng’s Estimated Net Worth is at least $1 billion
• DeepSeek’s flagship model, DeepSeek-V3, has 671 billion parameters

AI advancements are unpredictable, when no one was expecting to beat American AI giants, a small startup from China with just 200 employees, launched one of the most powerful AI LLM.

It’s impact was so huge that it wiped out $1 Trillion from the American Stock market. “YES A TRILLION WITH A T”

We gathered verified statistics about DeepSeek AI from various sources, like their research papers, documentation, and open-access info about the company.

In this post, I’m going to list the most important and by far shocking stats about DeepSeek AI. We’ll also talk about the allegations by OpenAI about data theft by DeepSeek and possible censorship that is discussed online.

Why you can trust BloggingCapital: We spend hours researching and testing every product or service we review, so you can be assured that you are buying the best. You can find out how we test & our editorial standards.

Table of Contents

• DeepSeek’s Popular AI App Is Explicitly Sending US Data to China
• DeepSeek’s Unethical Data Theft Can Make It Harder For TikTok To Sustain In the US Market
• China’s censorship is evident in DeepSeek AI Chatbot.
• DeepSeek Caught Hiding Information Due to Censorship
• DeepSeek’s app was ranked #1 in the Apple App Store on January 27, 2025, surpassing ChatGPT, which was in the #2 spot
• Deepseek’s Founder Liang Wenfeng’s Estimated Net Worth is at least $1 billion
• DeepSeek’s app has 2.6 million downloads by January 28, 2025
• DeepSeek is estimated to have 4-6 million active users worldwide.
• DeepSeek’s flagship model, DeepSeek-V3, has 671 billion parameters
• DeepSeek’s models are priced at $2.19 per million output tokens
• DeepSeek’s training cost is 1/10th and operational cost is 1/30th of ChatGPT.
• DeepSeek-R1 achieved a score of 79.8% on AIME 2024, slightly surpassing OpenAI-o1-1217.
• DeepSeek has a small team of about 200 employees compared to 5,300 employees in Open AI
• Other Important Facts About DeepSeek AI
• Things To Remember
• FAQ’s – Frequently Asked Questions
  • What is DeepSeek?
  • How efficient is DeepSeek’s AI model?
  • Has DeepSeek faced any security issues?
  • What impact has DeepSeek had on the tech industry?
• The Bottom Line

Disclosure: Please note that some of the links are affiliate links. When you purchase hosting using my affiliate link, they give me a small commission “at no additional cost to you.” I’ll earn a small commission. Which I promise to spend on complete trash 😂

• 💬 Updated Feb 1, 2025: Proofreading completed, fact-checked. Added insights from online forums.
• 💬 Published Feb 1, 2025: All the statistics were gathered, and verified. First draft was completed.

Let’s find some interesting deepseek stats. 👇

DeepSeek’s Popular AI App Is Explicitly Sending US Data to China

According to a report by Wired.com, the popular Chinese AI app is sending US data to China, potentially setting the stage for greater scrutiny.

Source:

1. Burgess, M., & Newman, L. H. (2025, January 27). DeepSeek’s popular AI app is explicitly sending US data to China. Wired. Retrieved from https://www.wired.com/story/deepseek-ai-china-privacy-data/

DeepSeek’s Unethical Data Theft Can Make It Harder For TikTok To Sustain In the US Market

After TikTok was banned for just 24 hours, which was reversed by President Donald Trump. A week after DeepSeek was launched, it was caught stealing data from ChatGPT and now sending US data to China.

These unethical practices can make it hard for TikTok as well to defend its position in the coming months.

Shivam, Founder Of BloggingCapital predicts this may trigger stronger action from the American side with tighter restrictions or possible ban for similar Chinese companies.

Source:

1. BloggingCapital. (Feb 1, 2025). DeepSeek Statistics. Retrieved [February 1, 2025], from https://bloggingcapital.com/deepseek-statistics/

China’s censorship is evident in DeepSeek AI Chatbot.

Several Reddit users believe that China’s DeepSeek is heavily censored. We analyzed this claim and conducted some tests on our own.

Posts from the chatgpt
community on Reddit

To test this we also asked some questions about china, here is the response.

Avoiding the First Question, Stopped replying after the second question

After multiple attempts, we found a similar avoidance tendency or the chatbot simply stopped responding.

Source:

1. DeepSeek. (n.d.). DeepSeek Chat conversation. Retrieved [February 1, 2025] Insert Month Day, Year], from https://chat.deepseek.com/a/chat/s/7485cbda-7014-4a30-8caf-2244374b39e5

DeepSeek Caught Hiding Information Due to Censorship

The user shared an interesting insight that DeepSeek intentionally hides information when asked questions that are censored.

Deepseek is only censored if you’re a slow reader
byu/Ambitious_Subject108 insingularity

Source:

1. u/singularity. (2025, Jan). DeepSeek is only censored if you’re a slow reader [Online forum post]. Reddit. https://www.reddit.com/r/singularity/comments/1hmugeq/deepseek_is_only_censored_if_youre_a_slow_reader/

DeepSeek’s app was ranked #1 in the Apple App Store on January 27, 2025, surpassing ChatGPT, which was in the #2 spot

This indicates a rapid adoption rate for the DeepSeek AI application, gaining popularity quickly since its release, and surpassing a major competitor. The app achieved this top ranking within a short time after its release. This success highlights the impact DeepSeek has had in the mobile AI space.

Source:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.

Deepseek’s Founder Liang Wenfeng’s Estimated Net Worth is at least $1 billion

Forbes estimates that DeepSeek founder, Liang Wenfeng, is likely worth at least $1 billion. This valuation is based on his ownership of the company and its potential, rather than solely on its current revenue.

According to Chinese corporate records, this estimate considers his ownership stake in DeepSeek, which is approximately 84%. Some analysts believe that his stake in DeepSeek alone could be worth $10 billion.

Source:

1. Forbes, Tognini, G. (2025, January 30). Here’s how much AI firm DeepSeek and its founder are worth.

DeepSeek’s app has 2.6 million downloads by January 28, 2025

This shows the immense interest in the DeepSeek AI application, with a very large number of people downloading the app in the very short time since its release. This indicates rapid user adoption and growing interest in DeepSeek AI. The number of downloads is likely to continue increasing given the app’s popularity.

While the exact Q1 download numbers are not available, the data strongly indicates substantial growth in app downloads, with more than 3 million in just the first half of January 2025, and over 10 million downloads by the end of January.

Source:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.

DeepSeek is estimated to have 4-6 million active users worldwide.

This shows the considerable reach of DeepSeek’s AI, with millions of users across the globe using it in the first month of 20255. This demonstrates the global impact of DeepSeek’s AI models. The user base is rapidly expanding, according to user analysis.

Source:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.

DeepSeek’s flagship model, DeepSeek-V3, has 671 billion parameters

This parameter scale makes it one of the largest AI models globally, demonstrating the company’s capacity for very complex models, comparable to top-tier Western large language models. The massive model size allows for more complex reasoning and processing. DeepSeek’s efficient architecture and reinforcement learning allow for these enormous models to be trained and run more efficiently.

Source:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.

DeepSeek’s models are priced at $2.19 per million output tokens

while similar models from OpenAI cost $60. The huge difference in price makes it a cost-effective alternative to other big AI platforms and is one of the most important statistics about DeepSeek. This is roughly 1/30th the price of OpenAI. DeepSeek’s pricing has spurred wider adoption and a price war in China.

Source:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.

DeepSeek’s training cost is 1/10th and operational cost is 1/30th of ChatGPT.

This proves that they’ve developed unique cost-saving measures, and can develop very advanced models on a comparatively limited budget. This is due to innovations like Multi-Head Latent Attention (MLA) and Mixture-of-Experts (MoE). DeepSeek-V3’s estimated training cost was only $5.5 million.

Source:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.

DeepSeek-R1 achieved a score of 79.8% on AIME 2024, slightly surpassing OpenAI-o1-1217.

On the MATH-500 benchmark, it attained 97.3%, also outperforming OpenAI’s model. The statistics show that DeepSeek-R1 model is on par with or better than some leading large language models in the world on some metrics, especially on math-related benchmarks. These scores demonstrate very strong performance on complex reasoning tasks. On LiveCodeBench, DeepSeek-R1 scored 65.9% compared to OpenAI-o1-1217’s 63.4%.

Source:

1. DeepSeek-AI. (2025). DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning.

DeepSeek has a small team of about 200 employees compared to 5,300 employees in Open AI

This is tiny, especially compared to OpenAI’s 5,300 employees, showing how incredibly efficient the DeepSeek organization is. This shows that DeepSeek can do a lot with comparatively little resources. The company emphasizes software-driven resource optimization.

Source:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.

Other Important Facts About DeepSeek AI

1. DeepSeek AI was founded in May 2023 by Liang Wenfeng, and it is backed by the High-Flyer hedge fund.
2. Most of DeepSeek’s models are released under the MIT license, allowing for free use, modification, and commercialization, to build a community of contributors.
3. DeepSeek relies on reinforcement learning (RL), which reduces the need for supervised fine-tuning, making models more efficient in terms of resources, and giving them an edge in reasoning and coding.
4. DeepSeek’s emergence has caused a global tech selloff and impacted the stock market, especially shares of Nvidia, whose stock declined by 13%.
5. DeepSeek employs Multi-Head Latent Attention (MLA) and Mixture-of-Experts (MoE) to optimize resource use.
6. A security breach was discovered where a database of more than one million records, including user data and API keys, was exposed on the internet.

Things To Remember

We urge you to exercise caution when using DeepSeek and, if best, to use the open-source GitHub repository to execute the system locally to safeguard your personal information and privacy.

FAQ’s – Frequently Asked Questions

What is DeepSeek?

DeepSeek is a Chinese AI startup founded in 2023, known for developing advanced language models that rival those of major tech companies.

How efficient is DeepSeek’s AI model?

DeepSeek’s latest model uses about one-tenth of the computing power compared to Meta’s Llama 3.1, potentially revolutionizing AI energy consumption.

Has DeepSeek faced any security issues?

Yes, in late January 2025, a database exposure revealed over 1 million log lines containing sensitive information, including chat history and secret keys.

What impact has DeepSeek had on the tech industry?

DeepSeek’s advancements have led to significant market reactions, including a $600 billion single-day drop in Nvidia’s stock, causing U.S. investors to reconsider their AI investments.

The Bottom Line

So these are all the verified stats regarding DeepSeek AI. I only wanted to include the most important and verified stats. For any conflict of interest, please contact us at shivam[@]bloggingcapital.com.

I hope you liked the new AI launched by Chinese Startup. We would love to hear your thoughts. HAIL AI. 🙂

References:

1. Flensted, T. (2025, January 28). DeepSeek AI Statistics and Facts (2025).
2. Agarwal, N. (2025, January 29). DeepSeek AI Statistics and Facts 2025 You Shouldn’t Miss.
3. Forbes, Tognini, G. (2025, January 30). Here’s how much AI firm DeepSeek and its founder are worth.
4. DeepSeek-AI. (2025). DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning.