This article was originally published on my blog and translated into English. For the latest version and future updates, please visit the original post: https://jaketao.com/language/en/ai-trends-2026
The capital markets' perspective on AI has been quite interesting lately. It began to cool off in the second half of last year, shifted to skepticism about the return on investment early this year, and has recently seen a resurgence of confidence. This is a stark change, and it has been like a roller coaster ride in just a few short months. While concerns remain, the most severe issues appear to be in the past. Regardless of how capital markets view it, AI development follows discernible patterns. By examining historical trends, we can make reasonable predictions about future outcomes, whether they are months or years away. As a professional in the tech industry, I would like to share some insights.
Models are stabilizing and AI applications are booming.
In recent years, many startups have focused on developing AI applications. However, they have encountered challenges as major updates to large language models have rendered their work obsolete. AI applications depend on optimizing and deepening the capabilities of large language models. If these models are upgraded too frequently --- and each upgrade brings a significant leap in capability --- companies will find it difficult to develop applications effectively.
Large language models have rapidly evolved, demonstrating capabilities that have advanced from those of an elementary school student to those of a middle school student and then a high school student in less than three years. It is reasonable to predict that they will soon become "college students," and these "college students" will most likely evolve into "platforms." Subsequent to this period of growth, the development will level off, which will be the ideal time to build AI applications. This trajectory mirrors the precedent set by the early stages of the internet and mobile internet. During the internet revolution of the 1990s, there was a strong interest in showcasing skills, but the period of significant growth and monetization did not occur until 1996 and 1997. It was not until after the dot-com bubble of 2000 that the industry truly began to flourish, and the major companies we recognize today all experienced significant growth during that period. The evolution of mobile internet technology has followed a similar trajectory. After the release of the iPhone in 2007, the market experienced a period of growth and development, leading to the emergence of various mobile applications in 2010. This development phase spanned approximately three years. A significant number of the companies with which we are currently familiar were established around 2010. This pattern is not solely attributable to the cycle of technological development; it is also influenced by the time required for the market to accept new innovations, for startup teams to enter the field, and for investors to adapt their strategies.
The significant popularity of OpenClaw at the beginning of the year demonstrated the market's demand for AI applications, a trend that has become increasingly evident recently as more and more AI applications are being implemented. I anticipate an imminent surge in vertical applications and the emergence of a new generation of startups.
A significant increase in the incorporation of artificial intelligence into business operations
Presently, the majority of AI applications are based on chat windows that operate on a question-and-answer model. While this model may not be the most efficient option, it is designed to address the complex user needs. However, as the field evolves, the demand for AI will become more specific, shifting from solving general problems at a broad level to addressing niche areas. This is analogous to the invention of the steam engine, which generated vast amounts of general-purpose power. However, the subsequent challenge of harnessing that power to solve practical problems required exploration by professionals in various fields.
It is generally accepted that companies can reduce costs by leveraging AI. However, the specific methods for achieving this are not always clear. At this stage, AI cannot function without human involvement. There is currently no indication that it will be able to operate independently in the future. Therefore, it is essential to establish a checkpoint at each work node. If we liken a company's operations to an assembly line, the most practical upgrade would be to divide the line into smaller segments, each with a control center staffed by an operator who verifies the AI's results before proceeding to the next step in the process. The process of segmenting and integrating systems is time-consuming and labor-intensive. However, once established, it will lead to a significant increase in work efficiency and a reduction in the number of employees required.
These enterprise-level services will become the subject of fierce competition among major tech companies. However, as was the case with website development in the past, the market is vast enough to accommodate all participants. Large companies will undoubtedly claim their share, but there will be ample opportunity for small businesses to thrive as well. This trend is expected to persist for several years. In the near future, many labor-intensive tasks will be replaced by AI, and companies will require only a few --- or even just one --- senior staff member to operate the control center.
Token Usage Experience Significant Growth, While Prices Show Decline
The rapid integration of AI into business processes is driving a substantial increase in token usage. This trend is already evident, and I anticipate that this year's surge is just the beginning. However, these measures are still inadequate. This phenomenon might be reminiscent of the early days of dial-up internet, when the amount of data being transferred was unimaginable. Could we have foreseen the immense surge in internet traffic and the significant increase in home bandwidth and internet speeds that we are currently experiencing?
Many people are unaware of the concept of tokens and assume that using AI is cost-free, similar to the internet. However, in reality, AI's token consumption often exceeds expectations. Each interaction and conversation uses hundreds or even thousands of tokens, and these tokens aren't inexpensive. The OpenClaw crayfish bot, which gained widespread attention last month, serves as a prime example. Many individuals enthusiastically implement these systems, subsequently encountering substantial financial surprises.
At present, the use of AI is primarily confined to the domain of "Q&A." Once consumer-facing (2C) applications experience significant growth and business-to-business (B2B) workflows are seamlessly integrated, token consumption will undergo exponential expansion. This should help explain why all the major tech companies are investing heavily in the construction of computing power centers regardless of cost. This will be the biggest growth driver for the next ten or even twenty years. The number of tokens essentially represents computing power, and computing power is fundamentally driven by electricity. Computing power is influenced by the performance of chips. However, electricity is an infrastructure component. In this regard, the U.S. power grid is significantly inferior to China's. However, the U.S. possesses many power-generation technologies independent of the grid, and these technologies will also flourish in the future.
In contrast to the fluctuations seen in a market economy, an increase in token usage will not result in a decrease in its price. As previously discussed, price is influenced by computing power and electricity costs. However, the deployment of data centers and the upgrading of power infrastructure will substantially reduce the cost of tokens, leading to a decrease in their price. In general, the cost of using AI should not increase significantly. This will stimulate the application market, as companies will adopt it to save substantial amounts of money.
From Horizontal to Vertical Development
AI is currently undergoing horizontal development, which refers to the continuous expansion of its general-purpose capabilities. Once it evolves into a platform, numerous vertical development opportunities will emerge. The concept is straightforward: once the platform stabilizes, AI applications are utilized to address a particular class of problems. As the number of applications addressing different problems increases, they will naturally become more specialized. The market determines the demand for these solutions. Customers require services that address their specific needs, rather than a comprehensive but versatile solution.
Presently, the degree to which AI is incorporated into other industries remains quite limited. As AI becomes integrated into business processes, entrepreneurs in various sectors will progressively identify issues that can be addressed by AI, resulting in the formation of more AaaS (AI as a Service) companies. These developments closely mirror the evolution of the internet.
Afterword
I had a sudden flash of insight this evening and would like to share my thoughts. This world is essentially a makeshift operation; for many things, once you grasp their essence, the path forward becomes crystal clear. However, this knowledge alone is not sufficient for success. External factors such as timing, motivation, and luck play a crucial role in determining outcomes. The unpredictability of our world is a source of fascination, but it also leaves us with a sense of regret.
Top comments (0)