Seeing a lot of discussions about how major CSPs (Cloud Service Providers) are selling tokens for open-source models with high gross margins. Actually, this aligns with what Liang Wenfeng mentioned during an internal exchange earlier: 'DeepSeek's API pricing logic is based on buying a batch of equipment from the market and recovering costs within ten months, which roughly corresponds to six times the profit.' Talked about this in detail earlier this month: The significance of CSPs deploying cost-effective open-source models and reselling tokens: https://((x.com))/qinbafrank/status/2074309065849884813?s=20 Future performance of major CSPs will redefine key metrics and values for AI commercialization: https://((x.com))/qinbafrank/status/2074754779755295164?s=20 Final thoughts summarized in the long post below. When major CSPs sell open-source tokens, they save a significant portion of revenue compared to reselling tokens for cutting-edge models (most of the revenue goes to the cutting-edge model manufacturers). It also has a higher premium than simply renting out GPUs. This premium comes from: self-developed ASIC computing; managed inference services; databases and vector search; object storage; networking and data transmission; agent runtime; security and identity; evaluation, logging, and monitoring; enterprise support services. Deploying cost-effective open-source models and reselling tokens is currently one of the most effective ways to boost AI business profit margins. Essentially, it commoditizes the 'model layer,' allowing CSPs to shift more value to the infrastructure + platform service layer (hosting, optimization, routing, governance, security). Naturally, gross margins will increase significantly. Of course, it was only mid-to-late June when major CSPs began large-scale deployment of open-source models, so the data isn't fully comprehensive yet. If the market later confirms that deploying open-source models yields better gross margins and faster cost recovery cycles for major CSPs, This could very well become the driving logic for the next wave of market trends.