DeepSeek Announces Major API Price Hike, Shifting from Low-Cost Competition to Monetization

Chinese AI startup DeepSeek, once renowned for its 'disruptive pricing' that shook the global AI market, officially announced on Thursday (the 6th) a price increase for its API services, explicitly stating that the hike will be 'rather substantial'.

This move comes just three weeks after the company introduced a 'peak and off-peak billing mechanism', indicating a strategic pivot from merely managing computational load to fundamentally repositioning product value and profit centers.

Usage 'Tsunami' Forces Pricing Reset

The backdrop to DeepSeek's price hike is the astonishing growth in model usage. According to data from the open-source tool OpenCode, the official version of DeepSeek V4 Flash has reached a daily token consumption of up to 8 trillion on the platform—surpassing the combined total of OpenRouter, a routing platform integrating over 400 models. On the Vercel platform, V4 Flash processes approximately 5.3 trillion tokens per week, making it the highest-volume model on the platform.

Regarding pricing logic, DeepSeek founder Liang Wenfeng explained during an investor meeting that API pricing is based on a profit model aiming to 'recover equipment costs within ten months after purchase'. He analyzed that at current price levels, user demand is highly insensitive to price changes, noting that 'even if prices double, there is little difference in token consumption'. This judgment serves as the direct theoretical basis for the current significant price increase.

In addition to massive traffic costs, DeepSeek's commercialization process is accelerating. The company is currently conducting large-scale fundraising and has begun preparations for an IPO as early as this year. Founder Liang Wenfeng now faces the first major challenge of balancing investor expectations, rapid market expansion, and the enormous capital required to build high-cost computing infrastructure.

On hardware demand, DeepSeek plans to construct a 1-gigawatt (GW) computing data center in Inner Mongolia, China. NVIDIA CEO Jensen Huang has estimated that the construction cost of a 1GW data center equipped with top-tier AI accelerators could reach as high as $50 billion.

Industry Impact and Market Expectations

DeepSeek's pricing adjustment is seen as a turning point in the Chinese AI market. In the past, the company's extremely low prices were viewed as a 'kill switch' for competitors, posing challenges to U.S. rivals like OpenAI and Anthropic. Robert Lea, Senior Industry Analyst at Bloomberg Intelligence, points out that while Chinese consumers still show strong demand for AI tools, smaller players like Baidu will face greater pressure as market leadership consolidates.

Market analysts believe that after DeepSeek's price increase, the overall industry price baseline will rise, giving competitors room to adjust their own pricing. For downstream developers relying on cheap APIs to build applications, future cost models will require large-scale recalculation.