Does AI really have a moat?

Does AI really have a moat?

2024 is the year of AI applications. Deep reasoning has become mainstream, AI agents have become a new trend, and multimodal models are emerging one after another. The iteration speed of AI products is fast, but the explosion moment has not yet arrived.

Computing power is not a moat. Since this year, there have been more and more voices that domestic and foreign large models cannot keep up. Relying solely on the improvement of computing power and parameter scale is difficult to surpass peers. Moreover, the training and operation of large models require large investments, and it is extremely difficult to achieve leading advantages and monopoly profits in the AI industry at this stage. The current situation of domestic large models is that the core functions are generally homogenized and there are many substitutes.

If AI applications want to break out of the encirclement, the keyword in the second half of the year has shifted from the “model layer” to the “application layer”. In the past year, domestic large model manufacturers have lacked distinctiveness in the model capabilities of general scenarios and the homogenization is serious. It is doubtful whether a breakthrough can be found by attacking from the perspective of application scenarios. Large models are the foundation, and killer applications are the key. However, there is still no spark in 2024. The overall AIGC is in the development stage and has deficiencies. AI mobile phones also show homogenization.

Commercialization cannot be rushed. The reason why AI applications are valued is that the market urgently needs to see AI technology transformed into a practical business model, and investors have high requirements for returns. Sequoia Capital has tilted its investment focus towards application development. The number of users of ByteDance’s Doubao is growing rapidly, but key indicators such as usage time are not ideal. The greater role of dialogue-based AI products is sample collection. Competitors are also in no hurry to commercialize it. Looking back at the development of the mobile Internet, AI technology has developed rapidly. Perhaps we can see the inflection point of the next iPhone era in 2025.