How is DeepSeek Making Money?

DeepSeek is making headlines for shaking up the AI industry, challenging established giants like OpenAI, Claude, and Meta with its cutting-edge models—all while offering them for free. The company first made waves with the release of DeepSeek V3, followed by its advanced reasoning model – DeepSeek R1, and now its vision model – Janus Pro 7B. These releases have not only rivaled models like GPT 4o, o1, Sonnet 3.5 but have also raised eyebrows due to their incredibly low training costs – just $5 million, a fraction of what competitors spend. This has sparked widespread speculation: how is DeepSeek making money while giving away its models for free? Here’s a closer look at the key points behind their disruptive strategy.

1. DeepSeek’s Core Business: A Quant Company

DeepSeek is primarily a quantitative trading company, specializing in building trading algorithms to generate profits. Their expertise in mathematics and optimization likely played a significant role in developing the DeepSeek R1 model. The company reportedly owns a significant number of GPUs, originally used for trading and mining purposes. DeepSeek appears to be a side project that leverages these GPUs efficiently, allowing them to train and run the model at a fraction of the cost compared to competitors.

2. Open-Source Strategy: Disrupting the AI Industry

By releasing V3 and R1 as open-source and open-weights, DeepSeek has disrupted the AI industry. This move challenges major players like OpenAI and Claude who have invested billions in proprietary AI models and infrastructure. The open-source nature of DeepSeek R1 allows anyone to reproduce and use the model, which has led to speculation that DeepSeek’s primary goal is not direct monetization but rather industry disruption and influence.

Also Read: How DeepSeek Trained AI 30 Times Cheaper?

3. Monetization Through API and Efficiency

• While the model itself is free, DeepSeek offers an API endpoint for running the model, which is reportedly extremely cheap. This low-cost API could attract a large user base, generating revenue through volume.
• DeepSeek’s efficiency in both training and inference (running the model) suggests they have developed innovative techniques to reduce costs. This efficiency could allow them to monetize the model at scale without needing to charge high prices.

4. Speculation on Hidden Resources

• Some industry experts, like Alexander Wang (CEO of Scale AI), speculate that DeepSeek may have access to more GPUs than they publicly admit. This could be due to export restrictions on cutting-edge chips from the U.S. to China, forcing DeepSeek to optimize their existing resources.
• If DeepSeek has a large number of GPUs, they could be using them to run the model at scale while keeping costs low, further enabling their ability to offer cheap API access.

5. Strategic Implications: A Wake-Up Call for the U.S.

DeepSeek’s success has raised concerns about the competitiveness of U.S. tech companies. The company’s ability to produce a state-of-the-art model at such a low cost has led to questions about the necessity of the massive investments being made by U.S. firms. Some analysts view DeepSeek’s move as a form of economic warfare, aimed at undermining the profitability of U.S. AI companies by setting a low price benchmark.

To know more, read this article!

6. The Power of Open Source

DeepSeek’s release of R1 is seen as a win for the open-source community. By making their model open-source, they have enabled smaller companies and researchers to compete with larger, proprietary AI systems. This aligns with the broader trend in the AI industry, where open-source models are increasingly seen as a way to democratize AI and foster innovation.

7. Long-Term Vision: Compute as the Ultimate Resource

Regardless of the cost of training, the real battle in AI is expected to be over compute resources. As AI models become more advanced, the amount of compute required for inference (running the model) will increase. DeepSeek’s efficiency in this area could give them a competitive edge in the long run.

Also Read:

• DeepSeek R1- OpenAI’s o1 Biggest Competitor is HERE!
• Building AI Application with DeepSeek-V3
• DeepSeek-V3 vs GPT-4o vs Llama 3.3 70B
• DeepSeek V3 vs GPT-4o: Which is Better?
• DeepSeek R1 vs OpenAI o1: Which One is Better?
• Kimi k1.5 Vs DeepSeek R1: Battle of the Best Chinese LLMs

Conclusion

DeepSeek’s monetization strategy appears to be multifaceted: leveraging their core expertise in quantitative trading, optimizing GPU usage, and offering low-cost API access. Their open-source approach not only disrupts the AI industry but also positions them as a key player in the global AI race. Whether this is a strategic move to challenge U.S. dominance or simply a gift to the open-source community, DeepSeek has undoubtedly changed the game in AI.

Stay tuned to Analytics Vidhya Blog for more such awesome content!

Nitika Sharma

Hello, I am Nitika, a tech-savvy Content Creator and Marketer. Creativity and learning new things come naturally to me. I have expertise in creating result-driven content strategies. I am well versed in SEO Management, Keyword Operations, Web Content Writing, Communication, Content Strategy, Editing, and Writing.