US stock AI investment enters the "second half": market volatility breaks the "blind buying also makes money" model, institutions seek "second-tier winners" in cloud computing and infrastructure.

Over the past year, artificial intelligence infrastructure has been a stable investment target in the US stock market, with many of the best-performing stocks in the S&P 500 index for 2025 falling into this category. However, as some stocks that were previously seen as winners began to decline, the volatility in this field also increased. Last week, NVIDIA Corporation (NVDA.US) CEO Huang Renxun stated that the new generation AI chip Rubin can be cooled solely by liquid cooling without relying on a water chiller. Cooling technology companies such as Trane Technologies plc (Trane) and Johnson Controls International plc (Johnson Controls) saw a sharp drop in their stock prices, as these two companies were considered beneficiaries of the approximately $475 billion in capital expenditures expected to be invested by Meta Platforms (META.US), Microsoft Corporation (GOOGL.US), Amazon.com, Inc. (AMZN.US), and Alphabet (GOOGL.US) in the next 12 months. On the other hand, SanDisk (SNDK.US) stock soared 37% last week and continued to rise on Monday, as Huang Renxun emphasized the demand for memory and storage. The stock achieved a 559% increase in 2025, making it the best-performing stock in the S&P 500 index, and then surged ahead by 64% in early 2026. This indicates that although the AI boom expanded in 2025, it is still a trend that is constantly evolving. As large tech companies continue to innovate, second and third-order derivative applications will emerge, but the companies and industries involved will not remain static. Eric Clark, portfolio manager of the Rational Dynamic Brands Fund, which holds a large stake in Amazon.com, Inc. and Microsoft Corporation, said: "Technology is advancing rapidly, things are always changing, and stocks related to liquid cooling certainly have risks." One way for investors to deal with market volatility is to hold stocks in multiple companies expected to benefit from AI-related capital expenditures. Last summer, Tortoise Capital Advisors launched an AI infrastructure ETF (TCAI.US), which holds stocks from multiple companies in areas such as heating and cooling, energy, storage, and building services. Since trading began on August 5th last year, the ETF has risen by 25%. Brian Kessens, Senior Portfolio Manager at Tortoise Capital, said: "We can't predict correctly every time, but we can certainly guess right a good portion of the time." He believes that market volatility caused by events like Huang Renxun's remarks may be a potential buying opportunity. "This may be an opportunity for us to realize that in the long run, these stocks are more appealing now because they will continue to benefit." Some are trying to diversify within the industry, betting on companies like power suppliers that have established mature businesses outside of AI and are likely to benefit from the major tech companies' significant spending. Brad Conger, Chief Investment Officer at Hirtle Callaghan, said: "Our investment strategy is usually to hold companies that may not have significant growth potential in the AI field but can participate in it. They will benefit from it, but this isn't their only advantage." The company holds shares in Vistra (VST.US), Constellation Energy (CEG.US), Amphenol (APH.US), and Eaton (ETN.US). The biggest question in AI infrastructure trading is whether large-scale data center operators will actually fulfill their promise of billions of dollars in capital expenditures. Any slowdown or reduction in spending could lead to market turmoil. Additionally, bottlenecks in areas such as water, electricity, memory, labor, and funding could hinder data center construction, exacerbating the volatility of related stocks. Matt Stucky, Chief Portfolio Manager of Stocks at Northwestern Mutual Wealth Management, said, "Healthy fundamentals are healthy fundamentals, and I don't think prolonging the window due to limitations in these fundamentals is necessarily the worst thing in the world. There is strong motivation to address product or supply chain-related issues." Of course, the AI industry as a whole is unlikely to disappear quickly. Therefore, even if there is a temporary slowdown in areas like data center construction, the industry still needs to build a large number of data centers for its continuous development. Kessens from Tortoise Capital said: "We like AI infrastructure, or rather, we like its second and third-order impacts because we believe it has long-term growth prospects, and we do not view AI as a bubble because large-scale data center operators are investing a large amount of capital and have ambitious plans."