Chinese AI startup DeepSeek has stirred things up again by sharing new details about the costs and revenues of its AI models. The Hangzhou-based company says its V3 and R1 models could theoretically have a daily cost-profit ratio of 545%, potentially making over $200 million a year. However, DeepSeek warned that actual earnings are much lower because of lower usage costs for V3, free web and app access, and discounted rates for developers during off-peak hours.
Low Costs, Big Impact
This is the first time DeepSeek has shared information about its profit margins for AI tasks like answering chatbot queries. This comes after DeepSeek revealed it trained its models with less than $6 million worth of Nvidia H800 chips, much less than what U.S. companies like OpenAI spend on AI training.
These claims have already impacted the AI market. U.S. AI stocks significantly dropped in January after DeepSeek’s R1 and V3 models became popular worldwide. The new cost-efficiency data could put more pressure on U.S. AI companies like Nvidia (NVDA), Microsoft (MSFT), and Google (GOOGL), which have invested billions in advanced chips. Investors might now question if such high spending is really needed, possibly leading to a rethinking of AI investment strategies. As DeepSeek shows, considerable AI progress can be made with lower costs, and the competition in AI could change significantly.
How Do AI Stocks Compare Against Each Other?
Let’s Observe how some of the more notable AI stocks compare to each other using Tipranks’ Comparison Tool. This tool helps us better assess a company’s financial standings and the market’s sentiments towards it.
