Recently, an event has occurred that has left the US AI community reeling. Despite the belief that China lags behind in chip technology and faces restrictions on acquiring top-tier AI chips, China has managed to develop an AI model that surpasses US counterparts.

This remarkable model is Deepseek, an AI large model launched by a small company in Hangzhou within just two months. It boasts capabilities superior to those of OpenAI's models.

More importantly, Deepseek achieves this with fewer chips, significantly less computing power than OpenAI, and training costs that are less than one-tenth of OpenAI's. The team behind Deepseek comprises graduates of Chinese universities, without any so-called "US study experiences."

How could such a team produce such an outstanding product? This achievement undoubtedly breaches the defenses the US had hoped to maintain.

Many view this as a stinging rebuke to the US AI ban. The US believed that blocking chips and talents would stall China's AI development, but this has proven to be wishful thinking.

So, what makes China's AI large model so powerful? Industry experts attribute this to Deepseek's innovative optimization strategies, including efficient load balancing, FP8 mixed-precision training, and advanced communication optimization technologies.

Furthermore, Deepseek adopts a different approach from OpenAI, focusing all efforts on a specific area, thereby obviating the need for massive computing power.

This underscores the fact that for AI large models, it's not necessary to obsessively stack AI chips. With the right methods and approaches, even a small number of chips can yield exceptional results.

The US has long believed that AI chips are fundamental and that by restricting them, it could maintain its global AI leadership. Given China's lagging chip technology, this seemed like a plausible strategy. However, it now appears to be nothing more than a fantasy.

Even Trump acknowledged that Deepseek served as a wake-up call for the US. It remains to be seen how the US will respond, as blocking chips has clearly failed to halt China's AI progress.