The Chinese AI “DeepSeek” has shaken the tech world. It has now been announced how much it is supposed to cost per day. And the costs of the Chinese model are significantly lower than those of the American one. Many investors are now wondering why ChatGPT is so expensive.
These are ChatGPT and DeepSeek:
- ChatGPT was introduced on November 30, 2022, sparking a hype around artificial intelligence. Many other companies, including Microsoft and Google, followed OpenAI.
- The Chinese start-up DeepSeek introduced its AI model DeepSeek in January 2025 as a competitor to ChatGPT. This literally caused a tremor on the US stock market. Companies like Nvidia lost billions in a single day.
As the English-language magazine Reuters reports, the rate of how much the Chinese model DeepSeek generates and especially costs per month has now been announced. And the costs are far below those of the American model OpenAI.
DeepSeek costs significantly less than ChatGPT
How much is DeepSeek supposed to cost? Reuters reports that DeepSeek estimates its total daily costs at $87,072 (around €83,500). This estimate is based on a rental price of $2 per hour (€1.92 per hour) for H800 chips.
The theoretical revenues of the AI are said to be $562,027. This means that the theoretical cost-benefit ratio of DeepSeek is 545%. In one year, this would amount to just over $200 million in revenue.
At the same time, however, DeepSeek limits itself: since many functions would be available for free, they would not be able to generate revenue from certain things. Therefore, actual revenues are likely to be significantly lower.
What does ChatGPT cost? The costs for ChatGPT, on the other hand, are much higher. The English-language magazine Futurism.com reports that the costs for ChatGPT are $700,000 per day. Although OpenAI now offers a subscription for $200 per month, the English-language magazine TechCrunch reports that OpenAI continues to operate at a loss and has not yet managed to generate profits.
Therefore, it is not surprising that this revelation has led investors to wonder why US companies spend billions on expensive hardware when they can achieve the same results with a lower investment. In contrast to the large expenditures of OpenAI, DeepSeek appears very economical.
Although DeepSeek has proven that it can operate at lower costs, its ultimate profitability will depend on how it monetizes its services in the future. Currently, the company is still far from the revenue projections of other industry giants, but its profitability is significantly higher. If the data provided by DeepSeek is accurate.
Criticism from the USA: Unsurprisingly, there has also been criticism from the USA. OpenAI, the company behind ChatGPT, stated that DeepSeek has stolen ChatGPT’s training data. There has long been the accusation that ChatGPT has not collected its data in a fair manner either. Or as the English-language magazine Vice writes: “OpenAI claims that DeepSeek plagiarized its plagiarism.”
Additionally, the CEO of Nvidia has recently spoken out, stating that Nvidia’s high losses were just a misunderstanding: For AI calculations still require fast chips, and companies will eventually buy them from Nvidia again.
Elon Musk believes that there are no longer any data from real people that can be used to train AI. And he is not the only one who believes this. Other major companies in the tech sector are already using synthetic data generated by AI: “We have reached the limit”: Elon Musk guarantees that there are no human data left to train AI
