According to DeepSeek, the digital services they offer boast an astonishing “cost-profit margin” of 545%. This claim was made by the Chinese startup in a GitHub post on a Saturday, outlining that only a select portion of its services are currently monetized. DeepSeek projected that if all users of its free AI models were to switch to a paid plan, the company could achieve a remarkable cost-profit margin of 545%. The post detailed that the company’s V3 and R1 models incurred a daily running cost of $87,072 at the end of February when leasing Nvidia chips at $2 per hour. Based on the pricing of the DeepSeek-R1 model, this usage could generate daily revenue of $562,027, translating to a cost-profit margin of 545%. However, it should be noted that these figures are based on specific assumptions and parameters, and might not accurately represent the current state of the company’s financials. DeepSeek operates on a system where during peak hours, its Nvidia H800 chips are utilized for inference tasks, while during off-peak hours, the focus shifts to research and training. These off-peak discounts, along with other cost-saving mechanisms, could potentially impact the company’s actual revenue. Despite these projections, questions have been raised regarding the feasibility and reliability of DeepSeek’s cost claims by industry leaders, researchers, and regulatory bodies.
Please rewrite the following text: er
Rewrite the text given above, ensuring to maintain the original content without any summarization or shortening. Also, please adhere to proper spelling rules and complete all sentences.
