The global "arms race" for computing power is well underway, with the competition shifting from "who has chips" to "who has factories and production capacity."
Since September, leading cloud service providers both domestically and internationally have continued to ramp up AIDC data center expansion: Microsoft plans to increase global data center capacity to 38GW by 2032, Alibaba Cloud announced at its Yunqi Conference a global data center scale exceeding 20GW by 2032, China Mobile Limited (ASX: 600941) disclosed at the 2026 Computing Power Conference that its total AIDC signed scale exceeds 7GW, and Google announced at least 13 billion euros of investment in a Finnish AIDC project. The expansion of computing infrastructure continues to materialize.
Every AI conversation, every image, and every video consumes Tokens behind the scenes. Tokens do not appear out of thin air—they require a physical "super Token factory": AIDC data centers that host GPUs, storage, servers, liquid cooling, switches, optical modules, and other hardware resources, packaged and converted into external computing power resources.
The ChiNext Computing Power ETF ChinaAMC (158041) anchors the computing infrastructure sector, focusing on data center AIDC construction and operations, computing power leasing, liquid-cooled servers, and other core computing power fields. Its index has the highest "Data Center (AIDC)" concept exposure in the entire market (91.03%), making it the sharpest tool for capturing computing infrastructure dividends, with explosive power far exceeding diversified but lower-purity indices.
Why computing infrastructure?
Policy, capital, and industry are all pulling in the same direction.
Over the past six months, the State Council has mentioned the computing power network five times. The recently issued "Opinions on Developing New Quality Productive Forces" mentioned optimizing the layout, construction, and operation of national data infrastructure, implementing projects like "East Data West Computing," building a nationally integrated computing power network, and strengthening efficient supply of computing power, algorithms, and data. On September 11, the discussion shifted from "whether to build" to "what to build and how to make it run." The meeting explicitly pointed out that the computing power network is the "foundational support for AI development," requiring coordination between computing and electricity, integration of computing and networks, and accelerated implementation of green power direct connection and source-grid-load-storage projects.
In less than a month, the National Development and Reform Commission led five coordination meetings, establishing a collaborative mechanism of "2 power grids + 3 telecom operators + N computing enterprises." Previously released "15th Five-Year" plans for the electronic information manufacturing industry and the information and communication industry emphasized building advanced computing ecosystems, setting a target of 9800 EFOPS in intelligent computing power by 2030, reducing PUE of new large and super-large computing facilities to below 1.2, and deploying ten-thousand-card and hundred-thousand-card intelligent computing clusters.
Further back, the Central Politburo meeting included the computing power network in the nation's "six networks" major infrastructure, alongside water networks and new-type power grids. The NDRC disclosed that new direct investment in the computing power network during the "15th Five-Year" period will be approximately 4 trillion yuan. The State Council also required deepening the "AI+" initiative and accelerating the deployment of super-large-scale intelligent computing clusters (AIDC).
Capital is accelerating its flow into computing infrastructure. The combined capital expenditure of the world's nine largest cloud providers for 2026 has been raised to approximately $830 billion, with annual growth increasing from 61% to 79%. Domestic actions are equally intense: Alibaba spent 67.7 billion yuan on capex in Q2, up 75% year-on-year; TENCENT (00700) spent 52.8 billion yuan, up 176% year-on-year. ByteDance raised its full-year capex plan to 200 billion yuan, about 25% above the original plan. Morgan Stanley estimates that China's AI-related capex from 2026 to 2030 will total approximately 8.5 trillion yuan.
Industry signals are also strengthening: demand is growing exponentially, supply continues to catch up, but the gap keeps widening. Current incremental computing demand is shifting from model training to model inference. China Telecom Research Institute estimates that China's annual Token consumption will reach 1 billion billion in 2026 and exceed 350 billion billion by 2030, with a compound annual growth rate approaching 12x. Over the next 2-3 years, China's computing demand is expected to grow nearly 10x annually on average. Data from CAICT shows that in Q1 2026, AI computing demand rose 417% year-on-year while supply grew 128%, with the computing gap continuing to widen. This supply-demand imbalance is driving up computing prices, enhancing industry pricing power for relevant manufacturers, transitioning from a "signing wave" to a "price increase wave."
These actions and this capital ultimately flow in the same direction—AIDC data centers, from construction to hardware, configuration, and operations, which are the targets covered by the ChiNext Computing Infrastructure Index (970083).
A purer index tool: filling the gap in A-share computing infrastructure investment tools
91.03% AIDC exposure—purity is sharpness. Those bullish on computing infrastructure previously lacked a sufficiently pure tool. Many existing "computing power" themed products are broad AI concepts mixing chips, models, applications, and software together. If you want to buy computing infrastructure, you end up with half a basket of other things.
What the ChiNext Computing Power Index does: focus on infrastructure. It selects 50 companies from ChiNext whose businesses must involve the four core hardware segments of computing, networking, storage, and operations, explicitly excluding software, media, and other "broad computing" components. Weight constraints are built into the methodology: single-stock cap of 10% for core segments like processors and computing equipment, and 3% for other segments, preventing any single company from dominating the index.
In terms of specific holdings: the index's Data Center AIDC concept exposure reaches 91.03%, the highest in the market. Additionally, optical module concept exposure is 53.51%, computing power leasing is 38.84%, and liquid-cooled server is 30.94%, with high weights in core components and core computing operation models. (Data source: Index popular concept exposure from iFinD, derived by comparing ChiNext Computing Power Index constituent stocks against 885887.TI Data Center (AIDC), 886033.TI Co-Packaged Optics (CPO), 886050.TI Computing Power Leasing, and 886044.TI Liquid-Cooled Server index constituents. Constituent businesses may have overlapping concepts. Data as of September 28, 2026.)
The highest AIDC concept exposure in the market means this index is the sharpest tool for capturing computing infrastructure dividends, with explosive power far exceeding diversified but lower-purity indices. Vertically along the AI industry chain, the index precisely positions itself in the third layer of the AI five-layer cake—infrastructure—locking onto the computing infrastructure track where AI capex is most concentrated and current earnings realization is most certain, filling a key gap in AI industry chain investment.
Prior gains have been fully digested, and the timing for positioning is approaching. The computing sector's earlier accumulated gains have been fully digested, with the index falling over 19% from its late-June high. Current trading congestion has noticeably decreased, and the sector adjustment is more of a phased fluctuation from profit-taking. (Data source: Wind, as of September 28, 2026.)
What better illustrates "digestion" than the price level is valuation. The index valuation has fallen back to a more cost-effective historical low range—PE, PS, and PB one-year percentiles are all below 2%, meaning the current price is cheaper than 98% of the past year, and the sector's sufficient adjustment has greatly digested earlier gains and bubbles. PEG = PE/net profit growth, i.e., how earnings growth digests valuation, to judge current cost-effectiveness. PEG is only 0.26, indicating expected growth far exceeds valuation, reflecting that constituent stocks' high growth has fully digested static valuation.
Earnings continue to materialize, with strong high-elasticity offensive characteristics. Beyond concepts, what's more worth watching is the financial statements. In the first half, the computing industry chain delivered high earnings growth, with the market shifting from valuation-driven to earnings-driven, and weight stock interim reports serving as the best footnote. The ChiNext Computing Power Index achieved 74.66% revenue growth and 189.48% net profit growth in H1 2026, ranking 2nd and 9th respectively among 199 indices tracked by ETFs in the entire market. (Data source: Wind, as of September 11, 2026.)
Why is computing infrastructure growth so strong? The underlying logic is not complex: high order visibility—a "orders first, revenue later" business model. The biggest difference between computing infrastructure and AI applications/AI software is that cloud provider capex directly converts into hardware orders, with orders preceding revenue recognition, making the earnings realization path clear and trackable.
Earnings have already materialized—not "expectations" but "landing." Constituent earnings materialize first—cloud providers and intelligent computing centers follow the industry rhythm of "invest CAPEX first, then form computing power, then train models, and finally launch applications," with hardware orders releasing before application revenue. Unlike software applications that need product refinement, user growth, payment conversion, and business model validation, computing hardware has a shorter monetization path: once cloud provider capex budgets are determined, they quickly convert into procurement orders for data centers and their components.
The index has strong offensive momentum—essentially the capital market pricing in advance for "high certainty, high growth, verifiable orders." As AI investment shifts from "storytelling" to "looking at earnings," while the AI application layer is still undergoing model iteration, product trial-and-error, and commercialization validation, capital is more willing to first allocate to the infrastructure segment with the clearest demand path. Therefore, the computing sector's early rally is an industrial consensus expressed through prices by global capital: in the AI industry chain, the application end has no clear winner yet, while the infrastructure end captures certainty-driven incremental growth first.
Global AI capex expansion, data center construction ramping up, computing power leasing seeing both volume and price increases—earnings materialize accordingly. This transmission chain is gradually showing up on financial statements.
The market also confirms this: the index has gained over 412% cumulatively since its base date of December 30, 2022, equivalent to an annualized return of 57%. High elasticity is the underlying characteristic of ChiNext growth assets. (Data source: Wind, as of September 28, 2026.)
————
Breaking down the AI industry chain, the chip layer tells the "sexiest" story, the model layer is the liveliest, and the application layer is closest to users. The computing infrastructure layer sits in the middle—perhaps inconspicuous in the past, but it is where AI productivity is truly "manufactured." The further AI advances, the more important computing infrastructure becomes, and the center of value gravity sinks downward. Stories about models and applications can be told for many years, but orders, earnings, and server rooms are visible. Computing infrastructure is moving from AI's "backdrop" to the "main battlefield."
The ChiNext Computing Power ETF ChinaAMC (158041) gives the market a sufficiently direct way to express its pricing of computing infrastructure. It also clarifies a previously vague question: when you are bullish on computing infrastructure, you no longer need to pick and choose among a pile of broad computing products.
Related ETF: ChiNext Computing Power ETF ChinaAMC (158041) tracks the ChiNext Computing Infrastructure Index, anchoring the infrastructure segment of the AI industry chain, focusing on data center (AIDC) construction and operations, computing power leasing, liquid-cooled servers, and other core computing fields. Its index has the highest "Data Center (AIDC)" concept exposure in the entire market (91.03%), filling the gap in A-share pure computing investment tools for investors.
Fund fee structure: Investors trade the above ETF on stock exchanges like buying and selling stocks. The main costs are brokerage commissions and fund operating expenses (management fee and custody fee at 0.50%/year and 0.10%/year respectively). When investors subscribe or redeem fund shares, the subscription/redemption agent may charge a commission of no more than 0.3%.
Risk disclosure: 1. The above fund is an equity fund, mainly investing in target index constituent stocks and alternative constituent stocks. Its expected risk and expected return are higher than mixed funds, bond funds, and money market funds, belonging to the medium-high risk (R4) category. Specific risk rating results are subject to the ratings provided by the fund manager and sales institutions. 2. The above fund is subject to main risks including deviation of target index returns from average stock market returns, target index volatility, and deviation of fund portfolio returns from target index returns. 3. Before investing in the above fund, investors should carefully read the fund's "Fund Contract," "Prospectus," "Product Information Summary," and other fund legal documents, fully understand the fund's risk-return characteristics and product features, and fully consider their own risk tolerance based on their investment targets, investment horizon, investment experience, and asset status. On the basis of understanding the product and suitability opinions, investors should make rational judgments and prudent investment decisions, independently bearing investment risks. 4. The fund manager does not guarantee that the above fund will be profitable or guarantee minimum returns. Past performance and net value levels of the above fund do not predict future performance, and the performance of other funds managed by the fund manager does not constitute a guarantee of the above fund's performance. 5. The fund manager reminds investors of the "buyer beware" principle in fund investment. After investors make investment decisions, investment risks arising from fund operating conditions, fund share listing trading price fluctuations, and fund net value changes are the responsibility of investors themselves. 6. The registration of the above fund by the China Securities Regulatory Commission does not indicate that it has made a substantive judgment or guarantee on the investment value, market prospects, and returns of the above fund, nor does it indicate that investing in the above fund is risk-free. 7. The above product is issued and managed by ChinaAMC, and distribution institutions do not bear responsibility for product investment, redemption, and risk management. 8. Markets carry risks, and investment requires caution.