AI penetration rate: Key indicator for grasping the market's "sense of position"

This week, some technology sectors have experienced adjustments. The focus of the market is on where this round of AI market is heading and how long it can continue. Here, we provide an important perspective and propose an indicator for grasping the market sentiment, which is the AI penetration rate. See the report for details: I. AI Penetration Rate: Key Indicator for Grasping Market Sentiment As a core indicator for judging the industry life cycle, the penetration rate can also help grasp the rhythm, position, and space of an industry market trend. The phase with the fastest increase in penetration rate is often the phase with the steepest market slope and the largest market imagination space. Once the penetration rate approaches saturation and the rate of increase starts to slow down, it means that the industry has passed the fastest development phase and entered a mature stage, and the market may also enter the top region. By comparing the current AI penetration rate with the Internet penetration rate during the dot-com era and the penetration rate of A-share main themes in previous cycles, we conclude that in terms of the current cycle, the AI penetration rate is still midway, not yet completing the phase with the fastest increase in rate; and compared to previous A-share main themes, there is still room for improvement in the current AI penetration rate. Firstly, the current penetration rate of personal generative AI in the United States is only at the level of Internet penetration in 1998, and has not yet completed the phase with the fastest increase in rate. According to World Bank data, the personal Internet penetration rate in the United States in 1998 was 30.1%. As of early 2026, according to the AI Index Report for 2026 from Stanford University's Human-Centered AI Institute (HAI), the overall penetration rate of generative AI for all Americans is 28.3%, just approaching the level of Internet penetration in 1998 and not yet completing the phase with the fastest increase in rate. Therefore, this round of AI cycle is still a distance away from the end, but has started to enter the second half, which is the stage of rapid diffusion of AI applications and continuous innovation in business models. Secondly, based on company research report keywords, we have calculated the proportion of companies within each industry that analysts believe are involved in "Internet+", "new energy+", or "AI+" related businesses to measure the penetration rate of industry trends at the level of listed companies. This indicator not only reflects the real level of business implementation but also demonstrates the extent to which market concepts and sentiments are spreading. In comparison, both from a holistic and industry-specific perspective, the current AI+ penetration rate still has room for improvement compared to the "Internet+" of 15 years ago and the "new energy+" of 22 years ago: Overall, in comparison to "Internet+" of 15 years ago and "new energy+" of 22 years ago, the current AI+ penetration rate still has room for improvement. The penetration rate of A-share AI+ has increased from 9% in 24 years to the current 21%, still lower than the 26% of "new energy+" in 22 years and the 30% of "Internet+" in 15 years. In various industries, compared to the previous "Internet+" trend, AI is still in its early stages of empowering most traditional industries. The penetration rate of AI in industries such as media, computers, electronics, automobiles, power equipment, and machinery has exceeded the level of the previous Internet era, but still remains relatively low in most traditional industries. Finally, one significant difference in this round of AI market compared to previous A-share market trends is that the previous growth trends, such as the "Internet+" from 2013-2015 and the "new energy+" from 2019-2021, were internal trends within the A-share market. However, this round of AI is a global resonance of technological revolution, the underlying DRIVE is the exponential growth in computational power demand, a trend that A-share market has not encountered before. Therefore, to grasp the trends and pace of this round of market, one needs to take a global perspective rather than focus solely on the domestic market. Comparing it to the dot-com market trend, it can be seen that the current AI market in the US is approaching the position of around June 1998: On one hand, a comparison of the trajectory of the Nasdaq index after the release of ChatGPT at the end of 2022 and after the release of Netscape at the end of 1994 shows that the current global AI market has many similarities to the dot-com bubble in terms of growth rate, slope, and rhythm. In comparison, the current AI market is at around 66% of the progress of the dot-com bubble stage in June 1998 in terms of time dimension and has achieved about 1/5 of the growth in space. On the other hand, as of May 22, 2026, the PE_TTM of the US technology sector is 35.9x, at the 69.7th percentile of the past ten years, and the Forward PE is 24.2x, at the 57.1st percentile of the past ten years. During the dot-com bubble period, the PE_TTM of US technology stocks could exceed 75x, and the forward P/E ratio could reach 45.5x. Compared to the dot-com bubble market, the current dynamic and forward P/E ratios of tech stocks are much lower than the peak in 2000, roughly corresponding to the level of May-June 1998. Therefore, looking at the current cycle as a whole, whether it is from the perspective of the penetration rate, market trend pace, or valuation comparison, the ongoing global resonance of the AI super cycle is still midway. II. June and July Remain Windows for Tech Industry Prosperity and Technical Intensive Catalysis In the short term, following some adjustments in the technology sectors and the relay of low-priced pro-cyclical sectors, the market is concerned about the sustainability of the AI market trend. We emphasize that the recent consolidation is more due to concerns about crowding on the denominator side, and the prosperity and industry trends on the numerator side are key to determining the sustainability of a major industrial trend in the market. Looking ahead, on the one hand, after experiencing recent consolidation, the crowding pressure on most segments of the technology sector has been somewhat relieved. On the other hand, looking forward to June and July, it is another window for the prosperity of the technology industry and intensive technological catalysis: Firstly, June is still a window for important conferences and industry developments in the domestic and international technology fields. The annual NVIDIA GTC conference, COMPUTEX conference, Microsoft Build developer conference, and Apple WWDC conference will be held intensively, with DeepSeek V4.1 focusing on optimizing multimodal capabilities expected to be launched in June, continuing to catalyze the North American computational power chain, large models, consumer electronics, and multimodal fields. Secondly, AI computational power demand, as the core engine driving this round of AI market, will also enter a key phase in June to verify and revise global computational power demand. First, optical communication leader Credo and core AI infrastructure provider Broadcom will release their performances next week, allowing the market to verify the sustainability of AI demand. Second, July will also see another phase of intensive catalysis for the financial performance of domestic and foreign technology companies. Third, according to historical experience, June will also enter a phase of intensive signal release for overseas CSP manufacturers and chip manufacturers in terms of order guidance and capacity planning, where the demand rhythm and performance profile of the North American computational power chain in 2027 will become clearer. Therefore, in the short term, with the concentration conferences and industry developments in the domestic and international technology fields in June and July, as well as the successive release of guidance for new rounds of performance and demand from listed companies, the highlights and focus of the global resonance are still in technology. Seen from the perspective of the ongoing global resonance, be it from penetration rate, market trend pace or valuation, the AI super cycle remains halfway through. Short-term concerns about consolidation caused by crowding need not be overly worrying; the sustainability of prosperity and industry trends is the key to the continuation of the market trend. III. Three Approaches to Allocation In terms of allocation, in the current situation where the consensus on prosperity at the numerator of the AI market trend is becoming more consolidated and pressures such as crowding and liquidity at the denominator are generally under control, we currently recommend sticking to directions with strong certainty of prosperity and focusing on the rotation and diffusion within the main theme rather than lowering positions or systematically shifting from high to low. We continue to recommend three approaches to allocation: 1) Play the long game and continue to stick to directions with the highest certainty of prosperity within the AI market; 2) Rotate and expand within strong main themes using our core approach of crowding, overseas mapping, and investor surveys that we have been tracking, focusing on exploring niche opportunities within the AI market; 3) Beyond AI, the overseas chain, especially in new energy and high-end manufacturing, is another main theme direction worth paying attention to. Specifically, the core conclusions on the directions: 1. For the North American computational power chain, represented by "light," attention needs to be raised once again. We have re-examined "light" based on four dimensions on May 26 and found hints of opportunities for "light" and the North American computational power chain, taking into account the comprehensive comparison of leading indicators, crowding, calendar effects, and valuation. On one hand, after June, the effectiveness of investing in prosperity is expected to improve again, and these varieties with high consensus on prosperity are likely to gather market consensus once again; on the other hand, following the diffusion of computational power to domestic production in the early stages, the leading indicators and crowding levels of the North American computational power chain have fallen back to low levels again, currently offering a good combination of odds and risks. 2. For the domestic semiconductor industry chain, while waiting for the digestion of crowding in the short term, it remains a direction with high consensus on prosperity and the greatest marginal changes in the medium to long term. Internally, it may be wise to focus on directions such as storage and CPU that still have room for further growth compared to overseas. 3. For the AIDC infrastructure sector, which currently has low crowding and strong consensus on market diffusion, it is necessary to increase its importance. 4. For midstream and downstream software applications, which currently have relatively low market consensus, but still being a direction with low crowding level within AI, and with the easing of anxieties about AI disruption, recent performances have been relatively good, including AI programming, data platforms, office software, and AIGC. 5. Beyond AI, the directions of low-end export chains and high-end manufacturing for continued growth diffusion also need to be given more attention. Combining crowding, prosperity, and industrial catalysis, it is important to pay attention to the diffusion of niche areas in new energy industry chains (such as lithium batteries, energy storage, photovoltaics, wind power, and power grids), as well as satellite communications, shipping, and engineering machinery. This article is sourced from the WeChat official account "XYSTRATEGY", authored by the Strategy Team of China Merchants Securities, edited by GMTEight: Chen Qiuda.