According to analysis, there are three main areas of focus. First, identify sectors where earnings expectations have already been revised upwards or have the potential for further upward revisions. This includes, but is not limited to, AI power infrastructure, machine tools/general automation, analog chips, and some software segments that were oversold due to the "HALO" trade (e.g., AI-generated web series). Second, within the broader technology sector, focus on areas with favorable positioning and a high density of near-term catalysts, such as commercial aerospace and humanoid robotics. Third, if there is a clear inflection point in the US-Iran situation and Taiwan Strait tensions, certain fundamentally sound metal sub-sectors, which have been significantly impacted by oil prices and liquidity since March, may demonstrate notable recovery potential.
**Event Overview** Since the renewed "TACO" rhetoric from former President Trump on April 8th, major global equity markets (primarily the US, China, South Korea, and Japan, which have high "AI" exposure) entered a Risk-On phase. Despite a gradual upward correction in oil prices diverging from "strong fundamentals," the impact of high oil prices on traditional economies has not yet been clearly reflected in high-frequency economic data. Meanwhile, the strength of AI infrastructure, both domestically and internationally, has been cross-verified and seen upward earnings revisions through financial reports and industry data, leading the market to price in this latter trend. Following the Labor Day holiday, signs of further easing in US-Iran tensions emerged, and navigation conditions in the Strait of Hormuz showed some improvement (Iran designated two shipping lanes). This led to a rapid decline in oil prices. Coupled with RMB appreciation and expectations of a Trump visit to China, strong micro-level industry fundamentals combined with supportive macro and liquidity conditions have driven a broad recovery in A-share risk appetite, moving beyond localized improvements. The average daily turnover for the entire A-share market has surpassed 3 trillion yuan, indicating a systemic "FOMO" (Fear Of Missing Out) sentiment among investors.
From a structural perspective, the period from April 8th to May 6th saw overseas computing power chains and domestic computing power/chip semiconductors leading the market in sequence. The former benefited from high earnings visibility, fitting the trading patterns of earnings season, while the latter validated the logic of "2026 potentially being the first year of significant volume for domestic computing power" through earnings and orders, representing a Davis Double play. However, as the A-share earnings season concluded and North American earnings reports entered their later stages, the expectation gaps for the highest-priority core sectors began to narrow, reducing their risk-reward attractiveness. With supportive macro conditions sustaining elevated market sentiment, sectors with lower earnings visibility but favorable positioning began to benefit from liquidity flows. This reflects portfolio rebalancing by existing funds (maintaining core positions while taking profits on overweight portions but remaining invested) and the entry of "FOMO" capital seeking new opportunities, objectively leading to a rebalancing of capital allocation.
In the final two trading days of the week, a significant catch-up rally was observed in small and micro-cap stocks relative to large-cap core holdings from a broad style perspective (conversely, the strength in small/micro-caps is also intertwined with abundant liquidity itself). Currently, market selection among lower-priced sectors is relatively dispersed. Initially, there was a steep catch-up rally within the computing power sector ("small optical" stocks). Broader tech sectors that are relatively low-priced and have catalysts, such as robotics and commercial aerospace, became active. Simultaneously, as the North American "HALO" trade reached an extreme, US software stock performance began to diverge, and A-share and H-share software stocks also saw some recovery. Additionally, some absolute low-priced, fundamentally bottoming domestic demand-oriented sectors attracted allocation from absolute return funds. Overall, the market is still in a rotational trial-and-error phase, with no single direction yet consolidating sufficient consensus.
Firstly, it is important to clarify that the medium-term uptrend in computing power hardware may not be over. During this North American earnings season, major cloud providers collectively revised their capital expenditure guidance upwards, with cloud growth generally beating expectations. Leading companies have substantial order backlogs (enhancing revenue certainty and validating AI commercialization). Furthermore, Apple, while reporting better-than-expected earnings, abandoned its net cash neutral target, suggesting the AI arms race acceleration cannot be easily disproven. Therefore, in the subsequent second and third quarters, barring major cuts in capital expenditures by large firms or financing difficulties, potential upward revisions to 2027 order and earnings expectations could drive the rally further at any time. Computing power remains the sector with the strongest fundamental logic in the entire market, implying that a "rotation from high to low" is unlikely to be systematic. However, in the short term (1-2 months), the gradual narrowing of expectation gaps objectively reduces the risk-reward appeal of computing power hardware. So, which sectors are more likely to generate alpha during this phase?
Two lines of thought: First, identify sectors where earnings expectations have been revised upwards but, due to various trading dynamics, have not been fully priced in by the market, primarily in derivative segments of AI infrastructure. Second, following the release of earnings risk, within the broader tech sector, areas with lower earnings visibility but possessing both favorable positioning and a high density of near-term catalysts are also worth attention. Additionally, if a clear inflection point emerges in the US-Iran situation and Taiwan Strait tensions, some fundamentally solid metal sub-sectors, which have been most impacted by oil prices and liquidity since March, may show strong recovery elasticity. The focus here is primarily on the first two directions:
**Category One: AI Power Infrastructure, Machine Tools/General Automation, Analog Chips, AI-Generated Web Series** Market pricing efficiency should be measured across multiple dimensions. In the A-share market, under conditions of ample liquidity, the peak of a major uptrend in core sectors can directly price in "future 2-3 year or even longer-term earnings levels * a valuation multiple." From this perspective, pricing efficiency is very high. However, this efficiency does not apply to every sector at all times. It may be due to linear extrapolation of pessimistic expectations caused by certain macro or industry uncertainties, triggering liquidity shocks (e.g., the mid-March sell-off, where software experienced a broad-based decline under the "HALO" narrative). Alternatively, it could be because a sector is not among the highest-consensus directions, and limited incoming capital often follows a pricing sequence (trading from the most prosperous to the next, with even "potential" prosperity being priced in during full bull markets). Furthermore, inherent information asymmetry leads to periodic lapses in capital consensus. "Early believers," based on industry information advantages and forward-looking judgment, buy on the left side to enjoy favorable risk-reward. "Late believers," as data continuously validates and positive price momentum creates a feedback loop, chase on the right side seeking probability, strengthening the trend momentum until all potential buyers FOMO in, concluding a complete trend cycle. Taking CPUs as an example, "the proliferation of AI Agents will increase CPU utilization" is not a new logic; it was discussed at the micro level at least since Q4 last year. However, what truly ignited the CPU rally (e.g., Intel in the US, and related partners or supporting chip companies in A-shares) was Intel explicitly stating "increased CPU-to-GPU ratio" in its earnings call and the boost from heightened Agent commercialization expectations driving stronger compute infrastructure demand, later validated by data.
From this viewpoint, attention can be paid to sectors where industry logic has reached a substantive inflection point/strengthened but, for various reasons, limited consensus has led to insufficient pricing. This includes, but is not limited to: 1) **Derivative segments of AI infrastructure.** Recent US earnings reports show the positive impact of AI infrastructure on earnings is spreading from chips/semiconductors to power and even some industrial segments. First, in power generation, earnings reports and communications from segment leaders like GEV, HWM, CAT, and BE have further revised upwards expectations for "power shortage" prosperity. This implies that for related A-share industrial chain companies, signals of volume growth (prosperity spillover from leaders) or price increases (especially in capacity-constrained segments) may become visible over the next 1-2 quarters. Diesel generators, gas turbines, SOFCs, and grid equipment related to power construction (also supported by domestic "six network" investment plans) are all worth continued attention. Second, for later-cycle semiconductor segments like analog chips and power semiconductors, Texas Instruments' report indicates the analog upcycle has officially begun. Besides a broad, mild global industrial recovery, demand from AI data center construction is a significant contributor. Similarly, power semiconductors (Wolfspeed Inc. has seen consecutive strong gains recently) are also benefiting. Additionally, expansion in advanced manufacturing sectors like AI liquid cooling, robotics, semiconductor equipment, new energy, and commercial aerospace is driving demand for general automation, making CNC machine tools noteworthy. 2) **Oversold software sub-sectors within the "HALO" trade.** Since February, the "HALO" trade has become a major narrative in global tech markets. While vibecoding fundamentally altering software development rules is one aspect, the broad-based sell-off in software, self-reinforcing narratives during price adjustments, and the dampening of sentiment for growth-oriented assets due to US-Iran tensions have amplified this negative feedback. In reality, the advent of the AI era does not mean the "total defeat" of software. Software companies can build enduring moats by deeply cultivating specific business scenarios, accumulating unique industry knowledge, and possessing vertical domain-specific data resources. After the extreme pricing-in of pessimistic expectations, software positioning has clearly improved. The IGV index in the US has recently shown internal divergence, with some "overcorrected" companies beginning to recover post-earnings. Within the software sector, focus can be placed on sub-sectors with clear industry inflection points, such as AI-generated web series. This segment is expected to achieve initial scale breakout in 2025, with commercialization likely accelerating this year, offering relatively high certainty within AIGC. Companies with high-quality IP reserves warrant close attention.
**Category Two: Focus on Commercial Aerospace and Humanoid Robotics** The industrial trends in commercial aerospace and humanoid robotics are certain, but near-term order and earnings visibility is low. Year-to-date (as of this week's close), their performance has lagged significantly behind the ChiNext and STAR markets. As catalyst density increases post-earnings season, they may become recipients of liquidity spillover from computing power and new capital inflows.
For **commercial aerospace**, the establishment of the Shanghai Commercial Aerospace Maritime Launch Technology Company in late April suggests industry development may accelerate. The sector faces a dense schedule of potential catalysts from May to July, including the next test flight of Starship V3 and the expected launch of domestic reusable rockets like Zhuque-3, which could boost sector attention. SpaceX plans to launch "Starlink V3" satellites in the second half of 2026, and progress in its pre-IPO work may release more positive industry signals, uplifting the sector. In the commercial aerospace rally from late last year to early this year, the core narrative was "orbital and frequency grabbing" under US-China tech competition, with launch costs not being the key factor. The start of a new rally may stem from favorable sector positioning and high catalyst density, but its strengthening will depend more on launch cost reduction and rocket reusability. Currently, the US Falcon 9 rocket is mature and lower-cost, while China is accelerating its catch-up. Rocket Lab USA, Inc.'s strong recent earnings and significant stock price increase on Friday indicate the North American launch services market is already booming. If Starship V3 achieves full reusability, launch costs could see an order-of-magnitude reduction—focus on SpaceX industrial chain targets. Domestically, reusable technology remains the core bottleneck limiting the scale development of commercial aerospace; this year is crucial for validation. If key reusable rockets like Zhuque-3 and Long March 10B succeed in reducing costs through reusability, the rally's potential will formally open up. At that point, value within the industrial chain will shift, warranting greater attention to satellite manufacturing, payloads, and communication application services.
For **humanoid robotics**, the annual strategic view is maintained. Starting this year, the sector's focus has shifted; its "thematic" nature has weakened, and vague upward revisions to long-term potential will have limited impact on the sector. The weight of medium-term fundamentals in investment considerations has increased. Tesla has announced the V3 robot is expected to debut mid-year and enter production in July-August. Therefore, pre-production factory audits may create order expectations for some companies. A strong trending rally will still rely on obtaining more credible information from industrial scale-up (e.g., unexpectedly high efficiency of humanoid robots from automakers in limited industrial trial settings like car factories, improving medium-term order visibility, and effective enhancements in model capabilities and economics enabling humanoid robots to adapt to more complex application scenarios).
**Summary** In summary, following the conclusion of the US and China earnings seasons, the short-term risk-reward profile for the core computing power direction has narrowed. The market may experience structural (rather than systematic) rotation from high to low. While maintaining core computing power positions, it may be appropriate to moderately explore opportunities in other directions, including: 1) Identifying sectors where earnings expectations have already been revised upwards or have potential for further upward revisions, such as AI power infrastructure, machine tools/general automation, analog chips, and some software segments oversold in the "HALO" trade (e.g., AI-generated web series). 2) Within the broader tech sector, focusing on areas with favorable positioning and high near-term catalyst density, such as commercial aerospace and humanoid robotics. 3) If a clear inflection point emerges in the US-Iran situation and Taiwan Strait tensions, certain fundamentally solid metal sub-sectors, most impacted by oil prices and liquidity since March, may demonstrate strong recovery elasticity.
Comments