The Chinese AI whizz, DeepSeek, recently lit up the news channels with claims that their AI models could be the modern-day money trees, albeit with a sprinkle of fine print.

DeepSeek didn't shy away from bragging about its financially successful online services in a particular tweet. The company announced a "cost profit margin" of 545%, a number derived from projected income.

This mammoth figure hails from a more detailed examination at the conclusion of a comprehensive GitHub post where DeepSeek elucidated its strategy to deliver "high throughput and low latency". The company expressed that if it was to bill every usage of its V3 and R1 models over a period of 24 hours with R1 pricing, DeepSeek's daily revenue would already be a staggering $562,027. On the other end, the cost of acquiring the necessary GPUs (graphic processing units) would be a plain $87,072.

While these calculations are super exciting, DeepSeek did concede that the actual revenue it makes is "significantly less" due to a mixed bag of factors - from nighttime discounts to lower pricing for V3 and limited monetization of services while keeping web and app access free.

But let's get real for a moment. If app and website access weren’t free and there were no occasional deals here and there, usage would naturally drop. So to say these calculations are speculative would be an understatement. It is more about sketching out the potential profit margins of DeepSeek's future rather than a peek into its existing financial health.

Regardless of the credibility of their figures, DeepSeek's bold step to publicly share these numbers is adding spice to the bigger discussion on AI costs and their prospective profitability. The limelight fell on DeepSeek at the start of this year with a new model that allegedly mirrored the performance of OpenAI's o1 on specific benchmarks. Managed on a lean budget and despite US trade limitations for Chinese companies to access powerful chips, DeepSeek really managed to shake up Wall Street, Tumbling tech stocks and raising questions among analysts about AI expenditure.

DeepSeek's creation wasn't only the talk of Wall Street. Its app knocked down OpenAI’s ChatGPT from the throne, temporarily, of Apple's App Store popularity list. It has since dropped off the broader rankings and currently perches at #6 in productivity, trailing behind ChatGPT, Grok and Google Gemini.