Will there be a bubble in AI? 🤔Where might it be? when? I saw this picture yesterday, it was made by Goldman Sachs. The comparison is the growth trend brought about by AI starting in 2023 (dark) and the rapid decline brought about by the bursting of the bubble after the rapid growth of the early boom of the Internet around
- A key question is, will this AI wave starting in 2023 also have a similar huge bubble? Why don't big data, cloud computing, and mobile Internet have bubbles like the Internet? My understanding is that these three stages are an extension of the Internet model to a certain extent. Network services will continue to become larger and broader, in various forms, from large equipment to small equipment, step by step, and everything will come naturally. Let’s briefly review the history of AI in the past two years. Starting from the explosion of OpenAI GPT chatbot, other traditional giants began to develop large models on their own. Nvidia continued to improve GPU performance, and basically confirmed that Scaling Law can work wonders. Model parameters are getting larger and larger, from simple question and answer to Function Calling to reasoning capabilities to Coding and agentic capabilities. This is probably such a development trajectory. Many excellent AI products and companies have been born, and consumers are quickly and willing to use some good AI products. At the same time, embodied intelligence is also developing. It is essentially an expanded scenario of AI + hardware. The fields of scientific research and literature also have the shadow of AI. It is basically a trend of full bloom and penetration, just like the grand occasion when everything was connected to the Internet after the emergence of the Internet. Many companies have grown a lot by relying on AI stories. Now it seems that Scaling Law is still effective. This is why we have recently seen various AI giants announce that they will continue to invest in the next few years. There will definitely be a wave of growth in the future. However, an interesting situation is that 2025 AI applications seem to have seen the homogenization of large models. Although they are lively, their performance is average from a payment perspective. In addition to the cost of Internet applications, AI products will also bear the cost of large model API calls as long as they are used. This requirement is actually quite high. How to stand out among homogeneous AI products? How to quickly follow up and integrate the continuously improving large model capabilities? You know, for those apps that are either huge or urgently needed, even if they don’t use AI, as long as they are well maintained, it will be difficult for them to lose users. What’s more, why don’t they use AI? Where are the opportunities for new products? I feel like I am looking for small blue oceans in the red ocean. So, my point of view is to be cautiously optimistic. Don’t think about ten-fold growth and returns. Is
- 5 times okay? Is 0.3 times okay? Work hard and steadily. If the AI bubble really comes, how will you survive? Not relying on financial expectations, but relying on VC money? So where does their money come from and where is it invested? It was also a serious issue at that time