<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/a123c57d-5fc7-46af-9b56-b20bac901e5c.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/a123c57d-5fc7-46af-9b56-b20bac901e5c.png"/>
On January 26, Jiang Shan, a fund manager at Invesco Great Wall, shared his views on AI computing power, AI applications, and robotics investment at the company's investment strategy meeting.
The investment representative summarized the key points as follows:
<ol>
<li>From the perspectives of model training paradigms and inference, several trends are quite certain.</li>
</ol>
In terms of computing power investment, first in the transmission segment, especially in optical communication, it will definitely be a segment of value inflation, with more new technologies and applications supporting the expansion of the entire market space.
There is also a definite increase in demand for storage in the process of needing longer context memory and larger KV Cache.
From the inference direction, first, ASICs may become an increasingly large market in the future. Second, under the trend of Agents, the tasks that CPUs need to complete will also increase significantly.
We can now observe some periodic price increase trends for CPUs. Looking ahead, similar to the new narratives of storage chips and CPUs in the past, as AI demand grows and changes become more frequent.
<ol>
<li>
By 2026, a very important point is that we will see many B-end applications gradually being replaced. Many application scenarios will be quickly replaced by AI.
</li>
<li>
Returning to investment targets and directions... a significant amount of value will be taken by leading large models themselves in the future.
</li>
</ol>
In terms of stocks, investable targets will become very scarce, similar to companies like Alibaba or ByteDance in the primary market that hold important leading large models, which will definitely have more long-term investment value.
...In the end, many of the targets we buy are actually damaged in this process.
It's like when the internal combustion engine came out, we invested in companies applying internal combustion engines but ended up buying a carriage company. It's a bit like that concept.
So from my perspective, I distinctly do not have a positive outlook on such companies. It's not that they are completely unviable, but I am not optimistic about most of these types of companies.
<ol>
<li>From a longer-term perspective, robotics is a brand new application of AI at the edge. Although short-term certainty is not particularly high, if we look at a time frame of over five years, it is a direction with very high certainty and significant space.</li>
</ol>
Jiang Shan, fund manager at Invesco Great Wall, has 12 years of experience in the securities and fund industry. He has worked at several institutions including Baoying Fund, Guotou Ruijin Fund, and Chang'an Fund, holding positions such as product manager, researcher, separate account investment manager, fund manager, and deputy director of the research department. He joined Invesco Great Wall Fund in December 2023 and has been serving as the fund manager of the mixed asset investment department since July 2024.
As of January 26, the total scale of the funds he manages reached 45.107 billion yuan. It is worth noting that in October 2025, his management scale was only 10.8 billion yuan,
This means that the management scale has grown more than four times in about three months.
It is also worth noting that the performance of the Invesco Great Wall Stable Return Mixed A Fund is outstanding, ranking 15th among actively managed equity funds with a return rate of 131.91% for the entire year of 2025.
As of January 26, the fund has accumulated nearly 390% in returns since its establishment nearly 11 years ago, with returns in various stages as follows: nearly 3 years 225%, nearly 1 year 115%, nearly 6 months 81%, nearly 3 months 14%, and nearly 1 month 6%. This indicates that there has been a recent performance adjustment.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/c42efb50-0ee0-4e56-afd1-34ac9d3c342c.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/c42efb50-0ee0-4e56-afd1-34ac9d3c342c.png"/>
We focus on its performance since August 2024, as Jiang Shan began managing the fund at this time. As seen in the chart below, the return since May of last year is approximately 152%.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/c16197c7-2012-44e5-80ae-cc862d26e926.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/c16197c7-2012-44e5-80ae-cc862d26e926.png"/>
Regarding views on AI investment, the official WeChat article from Invesco Great Wall Fund also elaborated on his perspective, which is consistent with the main theme of this conversation, but clearer and more straightforward.
The article states that he believes the key to computing power investment lies in identifying segments within the industrial chain that possess &#34;value inflation&#34; capabilities. Currently, two major trends can be focused on: first, the demand for high-speed data transmission and storage for large model training; second, the demand for high-performance integrated circuits and CPUs for large model inference.
Jiang Shan also emphasized that AI applications have gradually landed in multiple scenarios, including C-end chatbots and B-end automated processes (such as credit card bill assessment and mailing in the United States). In the future, quality targets will become increasingly scarce, such as companies holding leading large models. Additionally, from a perspective of over five years, the certainty of the robotics industry is very high.
The representative believes that he also proposed a core and sharp viewpoint in his speech: comparing revolutionary AI technologies (especially generative AI and large models) to &#34;internal combustion engines,&#34; while comparing software or service companies based on traditional technology architectures, workflows, and business models to &#34;carriages.&#34;
Jiang Shan hinted that one should not assume that a company dressed in the guise of &#34;AI applications&#34; can share in the AI dividends. Many traditional software companies (such as some enterprises in the A-share computer and media sectors) may have products that are essentially &#34;carriages&#34; that will be replaced by AI. Investing in such companies is not investing in the future of AI, but rather investing in a past that is about to be disrupted.
In the fund's quarterly report, Jiang Shan stated that he still focuses on AI computing power as the main layout direction, continues to hold positions in robotics and non-ferrous metals sectors, and has increased holdings in the power equipment and new energy sectors
Looking ahead to 2026, he believes that there is full confidence in A-share returns from the perspectives of industry fundamentals, liquidity, and policy. In terms of industry, AI remains the most exciting mainline.
As internet giants continue to increase their investments, the influence of AI will expand from the IDC industry to broader fields such as power grids, new energy, bulk electronic components, and even energy metals, creating a greater pull on the macro economy.
At the same time, the application side of AI is also expected to flourish in multiple areas: the maturity and revenue growth of B-end applications are promising, as AI will continuously open new scenarios and transform new industries; while the upgrade of smart devices and the progress of robotics will have epoch-making significance, pushing the AI industry into a new stage of positive interaction between application scenarios and capital expenditure.
The following are the highlights organized by the investment workbook representative (WeChat ID: touzizuoyeben), shared with everyone:
<h2>The ASIC market is growing larger, and there will be more new narratives for storage chips and CPUs</h2>
Host: The capital market often divides computing power into overseas computing power, domestic computing power, training computing power, and inference computing power. In the next stage of AI development, which specific track is the core?
Jiang Shan: I think the underlying question here is how to solve structural issues in the context of a positive outlook on computing power, and how to allocate positions within the computing power sector.
From our research, the study of computing power has entered its fourth year. Simply distinguishing between overseas and domestic, training and inference is too crude a classification method.
The more core issue is to study which links experience value inflation during the upward beta process, and which links may experience temporary supply-demand imbalances. These links can provide medium-term certainty and bring about temporary excess returns and stronger profit realization. This is a more core issue for me.
From the perspectives of model training paradigms and inference, there are several trends that are very certain.
First, in terms of training paradigms, there will be a greater emphasis on multimodality and the need for longer memory functions in the future. The computing power investments pointed to by these two directions will first be in the transmission links, especially in optical communication, which will definitely be a link of value inflation, with more new technologies and applications supporting the entire market space.
Second, in the process of needing longer contextual memory and larger KV Cache storage, the demand for storage is also certain to increase. Therefore, we see that whether in A-shares, U.S. stocks, or Japanese and Korean stock markets, these directions have created strong alpha over the past one or two quarters.
From the inference direction, first, ASIC may become an increasingly large market in the future. Second, under the trend of Agents, the tasks that CPUs need to complete will also increase.
We can now observe some periodic price increase trends for CPUs.
Additionally, looking forward, new narratives similar to past storage chips and CPUs, as AI demand grows and changes become more frequent, many upstream electronic bulk commodity links may transition from a state of blandness or even oversupply to a point of supply-demand balance.
This may be a more focused sub-direction in our future investment structure
<h2>In 2026, B-end AI Applications Accelerate Replacement</h2>
Host: Additionally, AI applications have been very popular recently. In the AI era, the scope of applications is broader than in the past, including pure software applications, edge applications, and future ones like superstar intelligence. I would like to ask two questions about this area: First, how do you view the stages of various application industries? Second, what are the investment opportunities for the rest of this year?
Jiang Shan: From our perspective, AI applications have gradually expanded in many scenarios. For example, many people may have more than one large model C-end software installed on their phones, and these entry points may become important personal assistants in the future.
I believe that C-end applications are more likely to gradually unfold in the form of chatbots. By 2026, a very important point is that we will see many B-end applications gradually replacing traditional methods.
The observable forms of these B-end applications may all involve replacing various forms of manual labor. This includes the gradual replacement of programmers in coding generation scenarios since last year; now many large companies may have a significant proportion of code generation completed by AI, and the replacement speed is still progressing very quickly.
Next year, we will also see other new scenarios, such as personal loans in the United States. If you have received a credit card bill in the U.S., you know that there are possibly hundreds of billions of dollars in personal credit card bills and loans, and the assessment and billing processes still heavily rely on manual labor, with about 3‰ to 5‰ of costs spent on assessment and manual work.
As one of the most profitable service industries in finance, similar application scenarios will be quickly replaced by AI.
This portion of value is the additional value that AI can provide to the real economy.
These are tangible ways that AI can improve labor productivity, gradually improving and replacing applications one scenario at a time. Therefore, I believe that 2026 will be a very important year for observing changes in the B-end.
<h2>Investment Targets Will Become Scarce, Headline Large Model Companies Have Greater Long-term Investment Value</h2>
Returning to investment targets and directions, as Mr. Zhong Wei just mentioned, a significant amount of value will be taken by the leading large models themselves.
In terms of stocks, investable targets will become very scarce, similar to the primary market, and companies like Alibaba or ByteDance that hold important leading large models will definitely have greater long-term investment value.
<h2>Not Very Optimistic About Most Application Companies</h2>
However, I speculate that what everyone wants to hear may not just be this; the core concern is whether A-share stocks in computing and media that are currently underperforming can be invested in.
Simply put, from the perspective of upstream enterprises' IT spending, AI spending will naturally crowd out other parts of IT spending. Especially in the current domestic environment, you may be leaders or executives in your respective companies; think about what expenditures would be squeezed out if your company were to invest an additional 20-30% of its budget in AI.
Ultimately, many of the targets we buy are actually harmed in this process. It's like when the internal combustion engine came out, we invested in companies applying internal combustion engines but ended up buying a carriage company. It's a bit like that concept
From my own perspective, I am quite distinctly not optimistic about companies like this. It's not that they are completely unviable, but I am not very optimistic about most of these types of companies.
<h2>Long-term focus on robotics, high certainty in the next 5 years</h2>
Host: What is the most promising sub-sector in 2026?
Jiang Shan: I would like to add one point. Robotics has also been a direction we have been focusing on for many years. Although this year we need to further confirm how the product capabilities and mass production rhythm are, there are still some observation points in the future that need to be verified. However, from a longer-term perspective, robotics is a brand new application of AI at the edge.
Although the short-term certainty is not particularly high, if we look at a time frame of more than five years, it is a direction with very high certainty and significant potential. This is also a direction we will spend a lot of energy researching for a considerable amount of time in the future.
Source: Investment Workbook Pro, Author: Class Representative
For more insights from industry leaders, please follow↓↓↓

Risk Warning and Disclaimer
The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at your own risk.

Jingshun Changcheng Fund Manager Jiang Shan stated at the investment strategy meeting that by 2026, B-end AI applications are expected to accelerate the replacement of traditional applications, and investment targets will become scarce, especially among leading large model companies. He pointed out that investments in computing power and robotics will have high certainty in the next five years. Although the certainty of robotics investment is relatively low in the short term, the long-term prospects are broad. He holds a cautious attitude towards investments in traditional internal combustion engine-related companies, believing that future investment opportunities will focus on companies with significant AI technology

Wallstreetcn

Invesco Great Wall Jiangshan: By 2026, B-end AI applications will accelerate replacement, investment targets are scarce, optimistic about leading companies in large models; the certainty of robots in the next 5 years is very high