DeepSeek, an AI-powered chatbot from China, has now become a serious competitor to OpenAI’s ChatGPT. One of the main advantages of DeepSeek lies in its significantly more efficient development costs compared to other AI models, including GPT-4 from OpenAI.

By spending only around 6 million USD (approximately IDR 97 billion) over a period of two months, DeepSeek managed to create two AI models, namely DeepSeek-V3 and DeepSeek-R1, which are claimed to be more cost-effective and efficient. This figure is much lower when compared to the development cost of ChatGPT-4, which reached 63 million USD (about IDR 1 trillion), according to a report from TeamGPT.

This substantial difference demonstrates that DeepSeek is not only capable of competing in terms of performance but also in terms of cost efficiency.

Efficiency Due to U.S. Chip Restrictions

China’s success in developing low-cost AI is largely attributed to the U.S. restrictions on advanced chip exports, such as the Nvidia H100 chip. In place of this, DeepSeek utilizes the Nvidia H800 chip, which offers lower performance but is more affordable.

These chip restrictions have pushed Chinese companies, including DeepSeek, to innovate and find more cost-effective development methods. One technique employed is "distillation," a process in which an AI model is trained to focus on specific tasks in a more efficient manner.

Chetan Puttagunta, a researcher at Benchmark General Partner, explained that this method allows the AI model to perform smaller tasks more intelligently and effectively.

Aravind Srinivas, CEO of Perplexity, a U.S.-based AI company, revealed that the chip restrictions have actually turned into an advantage for Chinese companies. "Many Chinese companies have been forced to find new strategies to overcome the chip restrictions, and as a result, they’ve managed to create far more efficient AI models," he said.

Tech World’s Response to DeepSeek

DeepSeek’s success in creating an efficient and high-performing AI model has drawn increasing attention from global tech leaders.

Satya Nadella, CEO of Microsoft, expressed his admiration for China’s ability to develop cost-effective AI. "I’m impressed by how they’ve managed to create open-source AI models that run efficiently while still delivering solid performance," Nadella stated in an interview.