AI: If you fear the mountains, what valleys are left?

```

The global technology industry is entering a critical period for verifying the cyclical recovery. U.S. tech giants reporting results above expectations, the Philadelphia Semiconductor Index setting a record of 18 consecutive gains, and Nvidia’s market cap returning to $5 trillion—all these signals indicate that the current rally is driven by earnings and constitutes a sustainable recovery, not just a sentiment rebound. The key issue now is: after the sharp rise in North America's computing power chain, which sectors still have a favorable risk-reward ratio?

Industrial Securities' strategy report presents a clear transmission path: The rally will diffuse from “low-valued segments within the North American computing power chain” to “domestic computing power and the AIDC industry chain,” and then to “mid- and downstream software applications.” This judgment is based on three logics: First, the low-valued segments within the North American computing power chain, which have the strongest prosperity consensus, sustained upward revisions in capital expenditure, and earnings growth forecasts generally between 60% and 130%, with more catalysts to come. Second, the domestic computing power and AIDC industry chain, which has been fully played out in U.S. stocks but not yet sufficiently reflected in A-shares, leaving considerable catch-up space. Third, mid- and downstream software applications, which are currently constrained by the earnings season’s demand for certainty, but after results, supported by continuous iteration of multi-modal technology, are expected to see a recovery window.

From a market capacity perspective, the overseas technology sector mapping window will begin in May to June, and the 40-day rolling performance gap between the TMT sector and the full A-share market is at only 2.66%, still significantly below the historical peak (around 10%). Therefore, concerns about “crowding” in the tech sector likely reflect internal structural issues rather than the core contradiction restricting the overall uptrend.

For investors, the key is to lay out positions in tiers: focus on low-valued opportunities within the strong North American computing power chain (such as fiber optic cables, liquid cooling, and power equipment), actively seize the domestic computing power and AIDC supply chain with the smoothest diffusion logic, and preemptively position for the elasticity in software applications’ recovery after results season.

US Tech Stocks Keep Hitting New Highs: Driven by Earnings Validation, Not Speculation

The strength of U.S. tech stocks in this round is mainly driven by the continued validation of earnings and industry trends, not merely by a restoration of risk appetite.

Recently, several milestone events have emerged in quick succession: The Philadelphia Semiconductor Index achieved 18 consecutive gains, a historic record; Nvidia’s market cap is back at $5 trillion; and leading players in different segments of the industrial chain like Amazon, Broadcom, TSMC, AMD, Micron, and Intel have all hit record highs, with many tech giants rising more than 50% since the end of March.

From the recently released Q1 reports, leaders across the industrial chain have generally exceeded expectations: companies such as Intel, TSMC, SK Hynix, Vertiv, and ASML all reported revenues, profits, and outlooks surpassing market expectations, as the AI-driven demand for the semiconductor supply chain is entering the phase of earnings validation.

Meanwhile, industry catalysts are arriving intensively: DeepSeek is seeking funding at a valuation of over $20 billion, with its V4 version announced and fully compatible with Huawei Ascend chips; Google has pledged to invest up to $40 billion in Anthropic and provide 5GW of computing power support; and CPU price hikes are spreading from the memory segment into more areas.

The above results and progress indicate that the main bright spot in the global industry is still technology, and the AI-driven demand for the semiconductor supply chain is entering an earnings validation window. In the coming period, intensive catalysts around global tech earnings and industry trends will persist, and the market focus remains on technology.

On the one hand, as U.S. tech giants continue to announce earnings reports from April to May, the sector’s prosperity is expected to be further verified. Of particular note, the four major U.S. cloud vendors will release their results this Thursday, with their AI ROI and capex guidance as the market’s main focus.

On the other hand, May to June marks the key window for overseas mapping of technology, which historically is also when domestic and foreign tech sectors have higher win rates. Major conferences like WWDC, COMPUTEX, and Google I/O developer conference will take place intensively. Historical data shows that the Philadelphia Semiconductor Index often delivers seasonal excess returns during this period, which also effectively translates to the A-share technology sector.

Crowding Is a Structural Issue, Not the End of the Cycle

Recently, there is a notable phenomenon in the A-share market: the leading gainers this week have mainly been resources and consumption sectors, reflecting the beginning of investor concern about overcrowding in tech and an attempt to rebalance towards cyclical sectors.

However, Industrial Securities strategy believes that crowding is most likely the main contradiction within the tech sector's internal structure, not the primary obstacle restricting the overall uptrend, for three reasons:

First, institutional consensus is concentrating towards the growth style. According to an Industrial Securities survey of over 180 core local institutional investors (April 21-24), compared with March when most focused on “stagflation trades” and “recession trades” triggered by geopolitical conflict, in April, the focus returned to industrial trends (57% in April vs 7% in March); and the consensus dominance of growth style has also strengthened significantly (16%→48%).

Second, seasonally, market catalysts will tilt towards technology growth. After Q1’s busy season, cyclical sectors’ excess returns usually converge in May-June, with industry catalysts concentrated in the tech growth direction reinforced by U.S. and Chinese resonance.

Third, TMT’s accumulated excess returns remain low. According to the rolling excess return metric, historic data shows that the top for TMT’s 40-day rolling relative returns versus the entire A-share index is typically around 10%, sometimes even reaching 20%-25% in some cycles. As of April 24, the rolling 40-day excess return for TMT vs all A-shares is just 2.66%, still significantly below historic peaks, so the sector as a whole has not accumulated much excess return.

Where Are the “Undervalued Areas”?

(1) North America Computing Power: Strongest Consensus, Focus on Low-Valued Segments

Currently, the North American computing power chain is still the strongest consensus sector within the AI supply chain, which also means that upcoming earnings seasons in China and the U.S. will continue to offer more catalysts.

Since the start of the year, net upward revisions to 2026 profit projections for AI sub-sectors are highly concentrated in network communication hardware benefiting from increased North American capital expenditure. In particular, fiber optic cables have seen the most significant upward revisions, with 2026 expected net profit growth surpassing 130%; for optical modules the revisions are about 21%, with growth over 120%; for switches, high-speed copper cables, PCB, etc., the upward revisions are in the 10%-16% range, with expected growth generally in the 60%-90% band, showing strong prosperity elasticity.

In terms of market performance, since the rebound from April 8, the sub-segments that have lagged behind in gains are worth close attention, including fiber optic cables, liquid cooling, power equipment, and switches. These links, given that their fundamental prosperity logic has not changed, present higher cost-effectiveness.

(2) Domestic Computing Power & AIDC: The Areas with the Smoothest Diffusion Logic

The domestic computing power and AIDC supporting supply chains are currently the most seamless routes for sectoral expansion, whether from overseas mapping, domestic catalysts, or market expectation perspectives.

Overseas, U.S. stocks have already demonstrated the diffusion path “from optical modules to semiconductors”: In early April, optical communication leaders like Applied Optoelectronics, Coherent, Lumentum led the rally; since mid-April, the baton has passed to semiconductor leaders like Intel, AMD, ARM. The A-share market has yet to fully reflect this logic, leaving notable catch-up potential.

Domestically, DeepSeek V4 has completed whole-chain training and inference adaptation on Huawei Ascend chips—Chinese computing power has moved from purely supporting model operation to large-scale inference validation for “long context, low latency, high throughput,” significantly boosting expectations for domestic computing power’s potential and incremental demand. At the same time, Intel's outperforming earnings validated the structural demand revaluation for CPUs due to large-scale AI agent deployment, currently the largest "expectation gap" in the AI supply chain.

In terms of market expectations, institutional surveys show the expansion path with strongest market consensus is highly concentrated on the domestic computing power and AIDC supporting supply chains.

In summary, as upward revisions for domestic computing power demand expectations continue and Token call volume grows explosively, special attention should be given to diffusion into two directions: first, the entire semiconductor supply chain, including leading domestic CPU, GPU, and memory chip makers, as well as material equipment, fabs, packaging/testing, EDA tools and other mid- and upstream segments; second, AIDC infrastructure, including power grids, computing power leasing, etc.

(3) Mid- & Downstream Software Applications: A Better Window After Results Season

The current tech rally has not yet been clearly transmitted to mid- and downstream “soft” areas such as computers and media, mainly due to three constraints:

First, U.S. software companies’ earnings have failed to alleviate concerns about how AI is disrupting business models; second, earnings in domestic mid- and downstream software apps mainly feature isolated Alpha highlights and have not yet built sector-wide confidence; third, DeepSeek V4 is still only a text model, lacks multi-modal capabilities, and is overall below market expectations.

With results season in full swing, it is recommended to prioritize the hardware areas with relatively high certainty of prosperity, including: AI edge hardware: consumer electronics, humanoid robots, satellite communications; beneficiaries of cloud service price hikes: IT services, cloud vendors.

After results season, as the market’s demand for prosperity certainty marginally diminishes, mid- and downstream software applications are expected to see a better recovery window. At that point, pay close attention to recovery opportunities in recently underperforming areas related to multi-modality, such as games, film and animation, and digital media.

Risk Disclosure and DisclaimerThe market has risks; investment should be made cautiously. This article does not constitute personal investment advice, nor does it take into account the individual investor's particular objectives, financial situation, or needs. Investors should consider whether any opinions, views, or conclusions in this article are appropriate for their particular circumstances. Investing based on this information is at your own risk. ```