A：The impact of AI on the U.S. economy is profound; its KPS already accounts for more than 2.5% of U.S. GDP and, through spillover effects such as chip procurement, influences East Asia.

A：The impact on China is mainly reflected in three aspects: First, export performance has been robust this year, with particularly significant benefits in the semiconductor and electronics supply chains; second, whether “token globalization” can become a medium‑term narrative is worth watching, as China’s advantages in infrastructure may help it achieve breakthroughs in this field; finally, domestic AI adoption is accelerating rapidly, and industry‑wide investment this year is expected to reach RMB 600 billion to RMB 1 trillion, contributing about 0.5 to 1 percentage point to the overall economy.

A：The Federal Reserve currently favors the ambiguous style of the Greenspan era, with diversified policy objectives and a heavy reliance on data. The Powell era may abandon certain practices of forward guidance, such as the dot plot. On the issue of raising or lowering interest rates, there is internal pressure to hike rates, but with inflation having eased, the Fed has yet to commit decisively to a rate increase. As a result, the Federal Reserve may refrain from taking significant action in the second half of the year.

A：Economic performance in the second half of the year is expected to exhibit a pattern of higher growth at both ends and lower growth in the middle. The third quarter will benefit from policy support, including fiscal spending, while inflation is projected to moderate, with the CPI remaining broadly stable. The stock market will exhibit a structural rally, with a slow, choppy bull trend prevailing. On the capital front, sector-specific opportunities will dominate. We recommend moderately managing portfolio exposure: while focusing on the AI theme, diversify into sectors such as securities firms and innovative drugs, and allocate to defensive assets like high-dividend stocks. There are limited opportunities in the bond market; investors can opt for stable coupon income while engaging in moderate tactical trading.

A：In this round of gold market moves, with few traders getting it right, our overall assessment has proven accurate. We had previously projected that the escalating tensions in Iran would affect inflation, the Federal Reserve’s interest-rate hikes, and the fiscal positions of numerous countries worldwide. Recently, Voss’s efforts to restore America’s credibility and the new wave of technologies brought about by the AI revolution have increased pressure on gold from investors. At present, the gold market is in a bottom-building phase. Although the most straightforward upward leg has already passed, the overarching trend remains intact, and there may be room for a modest correction going forward. It is recommended that individual and institutional investors reduce their gold allocation from the previous 5% to 10% to around 1% to 2% today.

A：In the second half of the year, the investment landscape in the stock market will exhibit a K-shaped divergence, characterized by pronounced disparities between the old economy and the new economy. At the macro level, this manifests as a K-shaped divergence, and at the stock level, it follows the same trend. Performance has diverged between the U.S. and Chinese markets: in the U.S., traditional consumer stocks have held up relatively well, whereas in China, consumer‑related sectors have lagged. In the second-half outlook, the key question is whether the K-shaped divergence between China and the United States will begin to narrow, and at what pace—both of which will shape the direction of China’s asset allocation strategy.

A：At present, several major divergences persist in the market; for instance, views such as de-dollarization and the appreciation of the renminbi driving capital inflows into Hong Kong stocks have yet to be substantiated. Forecasts that the Hang Seng Index could surpass 30,000 points and that consumer stocks would serve as a strategic investment have also failed to materialize. Our outlook diverges from the market consensus: we do not expect the U.S. dollar to trend persistently lower and, instead, remain optimistic about U.S. equities, having upgraded our U.S. equity outlook in our second-half forecast. Meanwhile, by analyzing historical performance, credit cycles, and other relevant data, we have identified a pronounced K-shaped divergence within both the A- share and Hong Kong stock markets, supported by fundamental evidence.

A：Both China and the United States exhibit a pronounced K-shaped divergence, with the bifurcation between technology and consumption stemming from differences in their respective funding sources. In the United States, funding is primarily provided by the private sector, giving rise to concerns about asset bubbles; in contrast, China’s government plays a more substantial role, exhibiting greater tolerance for short-term returns and costs, though this approach also poses certain long-term challenges. U.S. consumer spending remains relatively resilient, largely benefiting from the wealth effect, while China’s consumption—particularly in the real estate sector—has been weaker, primarily due to rising interest rates. Looking ahead to the second half of the year, determining which policy will be tightened first is crucial for shaping sector allocation, while also keeping a close eye on the impact of external demand pressures on policy and the role of fiscal stimulus in boosting domestic demand.

A：Yes, cash flow is a matter of great concern to everyone. The five major cloud providers have already exhausted their operating cash flow; without generating additional free cash flow, they will likely need to issue debt. Moreover, the source of funding is also a key perspective; currently, AI investment is primarily channeled by society at large—especially cloud providers, governments, or government-affiliated entities—but the sources of investment in the next phase remain uncertain.

A：In the short term, as AI’s positive impact on national strategy and economic growth outweighs its adverse effects on the job market, policies will continue to promote AI development, even though concerns may arise in the market regarding specific areas (such as dividend taxes).

A：At present, elevated real interest rates have made technology the preferred investment option, capable of offsetting high borrowing costs and delivering attractive returns. Although the Federal Reserve may raise interest rates further, with inflation largely peaking and the market having already anticipated and priced in some of the rate hikes, the impact on technology companies is likely to be limited.

A：Over the past few years, support from overseas markets—often referred to as “package deals”—has bolstered export resilience. Even amid trade tensions, emerging economies continue to build out their supply chains, while the transition to new energy is driving demand growth. This leaves a margin for achieving long-term growth targets, such as the 2035 goal; however, if exports do not weigh on overall growth, the fiscal position could remain stable.

A：Changes in the credit cycle can influence the performance of different asset classes: for instance, during a credit‑cycle contraction, stable-yield assets such as Hong Kong equities tend to be more attractive relative to other assets; while in the recovery phase of the credit cycle, a rebound in consumer spending tends to drive stronger performance in the Hang Seng Tech Index. As the credit cycle evolves, asset allocation should align with the direction of credit expansion and take into account the impact of interest-rate levels on cyclical assets.

A：The AI sector is currently experiencing cyclical fluctuations, and performance may slow down after the earnings season. Going forward, AI investments will be influenced by catalysts such as demand for computing power, the chip shortage, and model differentiation. Meanwhile, AI applications are diverging along B2B and B2C lines: B2B businesses are drawing significant attention for their computing-power needs and the use of domestically produced chips, while B2C businesses are being influenced by cost pressures and substitution effects.

A：Crowded trading does not alter an asset’s return characteristics, but it does influence its response to tail risk. Investors need to distinguish between good crowding and bad crowding and make decisions by weighing both the odds and the probability of success. Assets with high odds may offer substantial potential returns but entail significant holding costs, while assets with a higher probability of success could underperform due to underlying uncertainties and require catalysts to boost their likelihood of winning.

A：My second-half outlook is primarily grounded in an understanding of the diverging evolution of the China–U.S. credit cycles, from which I derive asset and sector allocation strategies. In terms of risk-reward ratios and win rates, I have illustrated the sectoral distribution across short-term market segments. While the consumer discretionary sector boasts strong risk-reward, its win rate remains insufficient; a policy shift could help boost its odds of success.

A：My understanding of future industry trends is based on a consideration of the intertwined effects of AI and population size. For example, although population decline may lead to overcrowding and concentration in central nodes, the development of AI could disrupt the job market and accelerate efficiency-driven substitution in specialized service sectors. At the same time, I remind myself to remain vigilant against a short-term mindset and to continuously reflect on evolving long-term industry trends.

A：AI and demographic factors will lead to a divergence between core and peripheral assets, with high-density population areas becoming more valued as they maintain the economic efficiency of public services. Moreover, AI will drive up the prices of certain standardized services, while non-standardized, customized services may also see price increases due to higher labor productivity, creating a new landscape of diverging prices between goods and professional services.

A：The widening income disparity is due to the fact that, in the AI era, returns to technology and capital have risen markedly, while labor’s share of output remains unchanged—this differs from the previous two technological revolutions. When computing costs remain constant, AI will seek to substitute human labor, thereby creating disparities in distribution.

A：Ordinary people can take the following measures: 1) Acquire skills and jobs that cannot be replaced by AI; 2) Earn customers’ trust, as emotional connections in interpersonal communication offer a degree of protection; 3) Master core node assets, non-standardized assets, and AI‑powered assets.

A：The brain–computer interface industry is currently in its early stages, with key applications including restoring bodily functions by decoding neural signals—such as Elon Musk’s Neuralink—and converting external information into signals that the brain can process—such as Xinzhi Da’s NorthBrain One. In terms of industry classification, the field is dominated by invasive and non-invasive approaches, with invasive brain–computer interfaces attracting greater attention due to their relatively lower safety profile. At present, some companies are working on language‑function restoration, but most remain in the development and clinical trial phases; visual‑restoration technologies are the ones progressing most rapidly toward commercialization.

A：Deep-brain electrodes, such as DBS electrodes, are comparable in safety and frequency to conventional electrodes, as they are designed to stimulate deeper, specific brain nuclei.

A：Semi-invasive electrodes (such as epidural and cortical strip electrodes) do not penetrate brain tissue. Epidural electrodes, like those produced by Boréas, offer relatively low signal quality, whereas cortical strip electrodes provide higher signal quality and better safety, approaching the performance of invasive electrodes while outperforming epidural ones.

A：Intravascular brain–computer interfaces use catheter-based techniques to deliver electrodes to the vicinity of the cerebral cortex for recording neural signals. Their primary advantage is a relatively high safety profile, while their limitations include a limited number of electrodes (compared with Neuralink’s devices) and the potential for blood flow to interfere with signal quality. Common consumer‑grade brain–computer interface products include sleep devices and meditation headbands, which can help promote mental relaxation. These are readily available online, though their effectiveness varies from person to person.

A：Ultrasound signals possess dual functions of readout and modulation, capable of influencing neural activity. They have already been employed in clinical cases to awaken patients in a vegetative state, demonstrating considerable potential; however, the technology remains in its early stages, and its long-term safety and efficacy still require rigorous validation.

A：In the medical field, DBS technology is used to treat Parkinson’s disease; Qiangnao Technology’s non-invasive brain–computer interface products are applied in the treatment of ADHD; and Mingshi Brain–Computer’s invasive brain–computer interface has achieved visual restoration and is slated to begin clinical trials in 2028, with prospects for market approval.

A：The market size has grown steadily over the past five years, and its growth rate is expected to accelerate to over 35% in the next decade, potentially reaching nearly RMB 100 billion by 2025. Healthcare remains the primary application domain, but the consumer‑end market, driven by rising penetration rates and advancing technological maturity, boasts substantial growth potential and may expand at an even faster pace.

A：In the United States, companies such as Neuralink, SambaNova, and Precision are advancing the development of brain–computer interface products in various forms. Neuralink’s device has entered human trials but has not yet received regulatory approval; SambaNova’s intracortical implant is in a pivotal clinical phase. Meanwhile, domestic firms including BoRuiKang, BrainCo, and Jieti Medical have made progress in both non-invasive and invasive BCI technologies, with BoRuiKang’s product already approved and in the early stages of commercialization.

A：Within one month of product approval, the product was assigned a medical insurance code, indicating that it is now being sold and billed in hospitals in full compliance with relevant regulations. Furthermore, from the payment perspective, this product has already been incorporated into Shanghai’s mutual‑benefit insurance scheme; ongoing monitoring of developments in other medical insurance reimbursement arrangements will be necessary.

A：Borui Kang initiated its IPO sponsorship in February of this year and completed the sponsorship in June, subsequently receiving acceptance of its application. If all goes according to plan, it is expected to go public before the end of this year. Its rapid listing process underscores Borui Kang’s representativeness and milestone significance in the brain–computer interface industry.

A：Non-invasive consumer‑grade solutions are advancing rapidly, though no blockbuster products have yet emerged; they remain in the early stages of commercial‑scale deployment. Semi‑invasive approaches are close behind, with Borey Kang’s accelerated product launch accelerating this trajectory. Meanwhile, invasive and interventional solutions are still in the clinical phase, representing the two slowest‑moving pathways.

A：The slower-than-expected progress of non‑invasive medical devices stems from the previous lack of clarity in national regulatory guidelines for such products. Relevant classification and definition policies were only issued at the end of June this year, placing most application scenarios under Class III medical device oversight. This has necessitated that companies revise their registration plans and clinical trial schedules to obtain marketing approval—including for brain–computer interface products—leading to an estimated commercialization timeline of one to two years.

A：In terms of its development trajectory, the concept of brain–computer interfaces was first proposed more than fifty years ago, but it has only entered the stage of human clinical trials in the past one to two years, with a relatively rapid pace of advancement. At the policy level, the state has designated it as a strategic emerging industry and, over the past year, has issued a series of supportive policy documents in rapid succession. In the secondary market, major national policies, the initiation of first-in-human clinical trials, and publicity campaigns by companies such as Neuralink have a significant impact on stock prices. However, in the first half of this year, despite the issuance of major policy announcements and progress in clinical trials, the stock price showed only a muted response, likely because the investment rationale has shifted, with the market placing greater emphasis on factors such as the company’s clinical‑to‑market execution and its technological‑barrier differentials.

A：In the short term, the extent to which Borui Kang’s products are covered by basic medical insurance and the inclusion of additional commercial insurance plans will directly impact the scaling of its commercialization. In the medium term, as more indications are approved and brain–computer interface technology transitions from clinical applications to productivity‑enhancing tools, financial performance will become a key area of focus. Over the long term, differentiated technological capabilities and the ability to expand into new use cases will determine the company’s competitive edge. Xiangyu Medical, a non-invasive medical device company, and Meihao Medical, an upstream player in the brain–computer interface sector, are both worth watching; investors can focus on the approval status of Xiangyu’s non-invasive brain–computer interface medical products and the investment opportunities along their respective supply chains.

A：NPO (Optoelectronic Co-Packaging) technology integrates optical-to-electrical and electrical-to-optical conversion directly onto the PCB within the switch or server. Compared with conventional optical modules, NPO is positioned much closer to the main chips—such as GPUs and switching ASICs—thereby reducing signal‑transmission loss, lowering power consumption, and enabling higher data‑transfer bandwidths with lower latency. Furthermore, the NPO employs a pluggable architecture, offering enhanced maintainability and flexibility, and is well-suited to high-bandwidth, low-latency applications in data center environments.

A：Moore’s Law has hit a bottleneck in the semiconductor technology field; simply relying on further miniaturization of individual GPU chips to smaller nanometer scales can no longer meet the demands of AI’s highly concurrent, clustered application scenarios. To address this challenge, the industry has begun leveraging multi-GPU fusion chips to enable parallel computing, thereby constructing compute nodes that resemble superchips. However, under these circumstances, inter-GPU data transfer has become a new bottleneck, leading to a sharp increase in the number of ports and a dramatic surge in power consumption and bandwidth requirements. Accordingly, NPOs and CPUs have emerged to address challenges such as port density, power consumption, bandwidth, and latency, driven by the critical need for efficient, low‑power interconnect solutions in data center deployment.

A：Thanks to its advantages in meeting data centers’ demands for high bandwidth and low latency, coupled with its currently very low market penetration, NPO technology—despite the mixed outlook for computing-power demand growth—offers both reliable near-term use cases and the potential for a future breakout, making it a compelling investment opportunity. Meanwhile, the advancement of NPO technologies will drive innovative applications of new processes and materials, such as the transition from conventional packaging to advanced packaging and material upgrades from indium phosphide to silicon photonics. Furthermore, both domestic and international tech giants such as Meta, Amazon, and Huawei are actively developing and advancing the application of NPO technology in data centers, which further validates its technical feasibility and market demand.

A：In the future, NPO technology will evolve toward higher data rates and greater integration; for example, 6.4‑Tbps NPO products have already entered the market, enabling upgrades to the rate and bandwidth of in‑rack optical interconnects. The core segments of the industrial chain include light-emitting chips, receiving chips, and DSP and electronic processing chips for optical signal processing. NPO will develop optical engine products by leveraging physical signal modulation and transmission media; among these, the optical engine commands the highest value and exhibits significant growth potential as customer demands evolve and technology advances. Meanwhile, external fiber-optic connectors, light sources, and other components will also see corresponding demand growth as NPO technology becomes more widespread. Ultimately, technologies such as NPO and CPUs will, based on a comprehensive evaluation of factors including yield, cost, power consumption, performance, functionality, and maintainability, as well as customer acceptance, collectively support the construction needs of data centers.

A：Due to the increased CPU clock speed and the growing number of channels, internal signal transmission must be implemented via optical fiber, and the light-source power will need to be boosted from 70 mW and 100 mW to 400 mW or even higher in the future. The light source employs a high-power emitter and is externally transmitted via an MPO (multi‑fiber‑optic connector).

A：Advanced packaging has become a national strategic direction because it is crucial to the future development of semiconductor and optical interconnection products. At present, TSMC's problem in CPU production is that the annual rate does not go up, resulting in higher costs and maintenance difficulties, but as long as the stability and yield are sufficient, its advantages will eventually appear.

A：Investment opportunities are concentrated among PIC vendors developing on-chip photonic engines for CPUs. These products require packaging by companies such as TSMC and incorporate silicon photonics, necessitating collaboration with silicon‑photonics wafer foundries like Tower Semiconductor or Zhongxing Hua Hong. Furthermore, the domestic‑substitution segment for ESC‑type electronic chips also holds substantial potential, with Chinese switch chip and optical interconnect module vendors poised to collaborate in developing independently controllable, self-reliant products.

A：The valuation of traditional Chinese equipment assets is low, but its market demand is undergoing significant changes on a global scale. In the past year, although traditional machinery assets such as construction machinery, ships, railway equipment, etc. have responded significantly to orders in a AI bull market, the valuation system still deviates greatly from advanced manufacturing AI assets. This is primarily because market demand for these traditional assets remains largely driven by domestic cyclical trends, while overlooking the emergence of a global supply inflection point and the growing overseas demand.

A：Take construction machinery as an example: Chinese equipment has been expanding overseas relatively early, and particularly since the Sino-U.S. trade tensions, as Chinese companies have increasingly sought to establish production facilities abroad, the penetration of construction machinery in international markets has continued to rise, with both profit and revenue contributions exceeding 50%. Meanwhile, data indicate that China’s equipment is steadily enhancing its global competitiveness: for instance, industrial robots now account for 50% of the global market, and the localization rate of certain automation products has surpassed the market share held by foreign brands, underscoring the robust competitiveness of Chinese equipment in both domestic and international markets.

A：The core of China’s domestic competitiveness is primarily reflected in the country’s comprehensive strengths in product quality, manufacturing capabilities, and supply chain systems, particularly as exemplified by industries such as construction machinery and shipbuilding. Among them, China’s share of global shipbuilding deliveries has reached 52%, far surpassing South Korea’s 28% and Japan’s 12%. By 2025–2026, China’s shipbuilding delivery volume is expected to remain above 50%, and the order backlog of China’s leading shipbuilders will continue to rank first globally. Furthermore, Chinese companies such as China Shipbuilding continue to secure orders in high-end sectors like VRCC LNG carriers, with their market share steadily expanding; the proportion of high‑end vessel types is expected to rise further.

A：The global revenue distribution among industrial automation companies reveals that Europe and North America remain the primary suppliers of high-end equipment manufacturing and exports. Take companies such as ABB, Schneider, Epson, and FANUC as examples: their revenue contribution in China remains relatively low, with Europe and North America continuing to hold the dominant position.

A：Supply rigidity is reflected in three aspects: first, long-term fixed investment. For heavy asset equipment such as gas turbine blades, a large amount of fixed asset investment and long-term depreciation are required to form production capacity; Second, the systematic and complete parts industry capacity, such as the machine tool injection molding industry, relies on the surrounding supply chain system to obtain cost advantages and response speed advantages. Third, engineer bonus, especially in lithium battery, photovoltaic, semiconductor and other industries, the iteration speed of automation equipment of Chinese enterprises exceeds that of foreign-funded enterprises.

A：Demand-side elasticity is primarily reflected in two aspects: AI-driven momentum and energy security reserves. The sudden surge in demand in the AI sector is driving up both the volume and prices of related equipment such as optical modules, PCBs, and semiconductor manufacturing equipment; meanwhile, the need for energy security reserves is shifting oil and gas capital expenditures toward investments in domestic energy security rather than profit‑driven ones, creating new overseas market demand for China’s traditional equipment.

A：Against the backdrop of globalization, Chinese equipment companies face a promising market outlook. On the one hand, they benefit from AI infrastructure-driven demand for power supply, thermal management, and production equipment; on the other hand, they profit from the cyclical upturn in commodities, which boosts overseas demand for mining machinery and construction equipment, as well as from strategic considerations of energy and supply-chain security aimed at building redundancy. This has opened up greater market demand and opportunities for Chinese equipment manufacturers.

A：Over the past year and more, despite fluctuations in the pace of interest-rate cuts, demand for construction machinery from Southeast Asian countries and Middle Eastern enterprises has continued to rise. This reflects that local needs for industrialization and energy self-sufficiency have surpassed traditional credit-driven profit‑seeking, becoming the primary driver of capital‑expenditure demand. Chinese companies’ market share in this sector has been steadily rising, now exceeding 30% globally, and reaching as high as 50% to 60% in Asia, Africa, and Latin America.

A：As oil and gas capital expenditures have become increasingly decoupled from oil prices in the wake of the U.S.-Russia-Ukraine conflict, companies have ramped up spending out of energy-security concerns, resulting in a shortage of related equipment and, consequently, boosting the global market share of domestic manufacturers. In the shipbuilding industry, overseas revenue and orders now account for more than 50% and 60%, respectively, whereas the general manufacturing sector lags behind. However, as the manufacturing cycle evolves, the pace of establishing and commissioning new plants has accelerated markedly across the globe, particularly in Southeast Asian countries, leading to a significant increase in exports of production‑oriented equipment over the past year.

A：In the past, we priced domestic companies based on domestic cycles, which resulted in relatively stringent valuation constraints. By contrast, for overseas capital‑goods firms, overall valuations tend to be higher when market share is substantial. As Chinese companies meet the needs of AI, energy autonomy and industrial autonomy, Chinese equipment is expected to usher in a double improvement in performance and valuation.

A：In the third quarter, the key factors under consideration primarily include the development of the AI industry, the potential changes brought about by the new Federal Reserve chair, and the performance of the domestic real estate and consumer markets. On the overseas front, economic growth is driven by countries with high AI adoption, while U.S. economic growth may weaken in the third quarter. In terms of liquidity, the U.S. dollar index could pull back, and inflation is expected to moderate steadily. Domestically, attention is focused on robust export performance, stable domestic demand, and the prospect that inflation has already peaked.

A：We recommend a slight overweight allocation to equities and a benchmark allocation to bonds. Within convertibles, we favor a modest overweight in both interest-rate‑sensitive and credit‑oriented segments, with a focus on high‑beta, speculative names that align with the technology theme. The technology sector is expected to remain active; it is important to closely monitor fundamentals and carefully select individual stocks. Hong Kong stocks are expected to rebound after a brief period of risk‑aversion-driven volatility, while crude oil prices are weighed down in the short term by easing geopolitical tensions but remain supported in the longer term by slow‑paced supply recovery and restocking demand. Gold, under pressure from constrained liquidity in the near term, may stage a catch‑up rally after a temporary pullback.