The claim that "Chinese AI contributes roughly 16% of global industry revenue, yet its value chain is severely undervalued" is gaining traction.
The AI bull market does not lack signals; capital flows towards wherever profits are moving.
On July 8th, the Hang Seng Tech Index in Hong Kong surged 4.97% in a single day, with Alibaba Group Holding Ltd (NYSE: BABA) soaring 12.21%, marking its largest single-day gain since September 2025.
Market consensus suggests the sudden surge for the Hang Seng Tech Index may be linked to the latest weekly report from Michael Wilson, Morgan Stanley's Chief U.S. Strategist.
In his report, he advised investors to reduce their overweight positions in the semiconductor sector and instead focus on hyperscale cloud service providers. He posits that AI investors are shifting their focus from chasing upstream hardware to concentrating on the commercial returns from downstream cloud services.
Among the top ten constituents of the Hang Seng Tech Index, companies like Alibaba, Tencent Holdings Ltd (OTC: TCEHY), Xiaomi Corp (OTC: XIACF), and Baidu Inc (NASDAQ: BIDU) are all deeply involved in AI models, cloud computing, and the AI application ecosystem. Following the investment bank's spotlight, these stocks quickly attracted significant capital inflows.
A day later, Goldman Sachs also made its move. On July 9th, it released a report titled "Long China AI Value Chain."
As the title suggests, Goldman Sachs argues that China's AI-related companies contribute about 16% of global AI industry revenue, yet the value of this industrial chain is severely undervalued, presenting an excellent opportunity for investment. The market interprets this as Goldman Sachs hinting that some capital could rotate from high-valuation markets like South Korea and Taiwan, China, into China's AI industry chain.
Once again, following the "prophecies" of the investment banking giants, the three major A-share indices surged on July 9th, with the STAR 50 Index jumping 6.67%.
Upon closer examination of these two investment banks' supportive actions, a long-standing hypothesis of "AI profit migration" comes to the fore.
Understanding the AI Profit Migration Chain
In March of this year, NVIDIA CEO Jensen Huang used an analogy, describing the AI architecture as a five-layer cake, from bottom to top: Energy, Chips, Infrastructure, Large Models, and AI Applications.
Industry insiders have further decomposed this into a three-tier industry chain: upstream, midstream, and downstream.
The upstream provides the physical hardware support for computing power, data storage, and transmission, such as AI chips, optical modules/components, AI servers, and networking equipment.
The midstream consists of the computing power network and model construction layer, which includes large model developers, cloud computing providers, and computing center operators.
The downstream encompasses AI agents, vertical application integration (AI+), and large model applications.
Historically, during different stages of industrial development, profits continuously concentrate in the most scarce and most price-negotiable segments. Capital market funds chase these profits, flowing from one segment to another, always moving into the industry within the AI chain with the highest profitability.
This forms the "Profit Migration Hypothesis." In other words, the protagonists of the AI bull market are not fixed; they change as the industry's center of gravity shifts.
So, where is this profit transmission chain currently? It has naturally reached the already richly rewarded AI hardware and AI infrastructure industries.
The most prominent example is NVIDIA's GPUs. Driven by explosive demand for AI infrastructure and a near-monopoly market environment, NVIDIA reported a Q1 FY2027 net profit of $58.32 billion, a staggering 211% year-over-year increase, leading the global semiconductor industry in profitability.
Subsequently, profits "migrated" to a series of AI hardware products within NVIDIA's ecosystem. These previously unremarkable components also saw their profit margins skyrocket against a backdrop of supply-demand imbalances.
Recently, prices for some AI data center-related optical communication products surged 400%-650%, certain 1.6T optical module products increased by 70%, and HBM4 memory prices quadrupled. Related stocks thus became the next objects of capital pursuit after NVIDIA, with investors holding them "standing in the light and walking in the chips."
The story doesn't end there. Following the HBM memory price surge, the market continued its reasoning: having NVIDIA's GPUs and high-bandwidth memory isn't enough. All the nodes and commodities needed to build AI data centers could become new profit growth points.
Thus, starting this year, areas including AI power supply, land for computing centers, and liquid cooling have all garnered attention from the capital markets. The "shovel sellers" of the AI era have expanded from chip suppliers like NVIDIA to the entire hardware infrastructure layer.
However, the capital market soon posed a new question: If AI infrastructure construction continues to expand, who will ultimately pay for this ever-increasing computing power?
Consequently, capital began shifting its gaze from the "shovel sellers" to the "shovel users," with focus gradually moving from the hardware infrastructure layer to cloud computing providers, large model companies, and AI application developers.
From Optics to Cloud to Models?
According to Morgan Stanley's analysis, cloud computing providers have spent massive sums on procuring computing power over the past few years, acting as the paymasters. Now, the market is beginning to re-evaluate AI returns, discovering that some cloud providers have already transformed their previous massive investments into growth in revenue and profit.
J.P. Morgan released a research note on July 8th stating that Alibaba's upcoming Q1 FY2027 results are likely to exceed expectations, based on three key points: rapid narrowing of investment losses in food delivery and instant retail, further acceleration in cloud business revenue growth with quarter-over-quarter margin improvement, and a quarter-over-quarter decline in losses from other businesses after excluding Spring Festival customer acquisition expenses.
The report estimates Alibaba's Q1 FY2027 cloud revenue grew approximately 45% year-over-year, accelerating from the previous quarter's 38%, with full-year growth expected at 47%. It also noted that in Q4 FY2026, AI-related products already accounted for about 30% of Alibaba's external cloud revenue, with current demand continuing to outstrip supply capacity.
Regarding profitability, Morgan Stanley also calculated that Alibaba Cloud's cloud EBITA margin improved from 9.1% last quarter to 12%.
The investment banks' optimistic forecasts for Alibaba send a signal to the capital market: cloud computing providers, as suppliers of AI large model computing infrastructure, are beginning to realize profits.
With revenue and profits rising, a corresponding increase in related stock prices follows naturally.
Following cloud computing, the capital market simultaneously turned its attention to large model developers. The two share a symbiotic technological relationship; cloud computing provides the computing infrastructure for large models, while the quality of the models determines whether cloud providers can sell tokens effectively.
So, are large model companies profitable yet?
Looking first at the U.S. leader, Anthropic. With the popularity of Claude Code, it achieved profitability first. It is projected to record an adjusted operating profit of $559 million in Q2 2026 and expects GAAP EBIT to exceed $1 billion in Q3.
Turning to China, as of March 2026, the annual recurring revenue (ARR) for Zhipu AI's MaaS (Model-as-a-Service) API platform reached approximately 1.7 billion yuan, a 60-fold increase over the past 12 months, with gross margin improving nearly fivefold to 18.9%, indicating significant commercialization efficiency improvements.
J.P. Morgan's research expects Zhipu AI to generate revenue of 3.192 billion yuan in 2026, a year-over-year increase of 340.6%, and believes the company will achieve profitability by 2029.
Using them as references suggests that leading large model enterprises in both China and the U.S. have taken significant steps forward in commercialization and profitability.
Furthermore, another application platform also in the downstream is experiencing explosive growth: Agent software.
According to the "AI Agent Empowers Industry Decision-Making: Trends and Practice White Paper" published by CCID Consulting, the market size for enterprise-level AI agents in China is expected to reach 44.9 billion yuan this year, potentially exceeding 332 billion yuan by 2029, with a compound annual growth rate as high as 107%.
If the AI bull market of the past three years saw capital chasing the "shovel sellers," today, the capital market is searching for companies that can genuinely use AI to generate profits.
What Should Individual Investors Do?
A more practical question for investors is this: if the "AI profit migration" hypothesis holds, does it imply that the AI sector still holds significant potential for the latter half of the year?
Reviewing the mid-year outlook reports from several major brokerages, the answer appears affirmative.
Among the high-frequency keywords in these reports, "AI" appears approximately 355 times and "technology" about 108 times, far exceeding other categories. Terms like "computing power" (34 times), "semiconductor" (33 times), "crowded" (55 times), and "valuation" (41 times) also feature, with AI remaining the strongest investment theme.
Simultaneously, the way some institutions discuss AI has changed.
Previously, what excited the market were technological breakthroughs, model capabilities, and industrial imagination. Now, when looking ahead to the second half, institutions like CITIC Securities Co., Ltd. (SHA: 600030), China Merchants Securities Co., Ltd. (SHA: 600999), and China International Capital Corporation Ltd (SHA: 601995) repeatedly mention key terms: revenue, profit, ROIC, and cash flow.
To some extent, this echoes the core viewpoint of the "AI profit migration" hypothesis: follow the flow of profits within the industry chain to find the next batch of companies capable of capturing those profits.
However, from a practical trading perspective, several institutions believe the market in the second half will be "much harder to play," potentially entering a complex phase with increased volatility and declining risk-reward ratios.
In other words, the bull run will likely continue, but its path will be more tortuous.
For individual investors, while profits may migrate along the AI industry chain, the decision of whether to follow that migration themselves tests both their strategic resolve and tactical caution.
Comments