AI market faces a critical test! During the internet bubble period, star analysts recommend reducing holdings of cloud giants and focusing on the semiconductor sector.

In the recent cooling trend of global technology stocks, Dan Niles, a well-known semiconductor analyst from the Internet bubble era and founder of Niles Investment Management, warned investors that artificial intelligence (AI) trading is about to hit a "speed bump." He believes that smart money should flow into the actual spending area of AI funds - the semiconductor industry, rather than the large-scale cloud computing companies that are currently receiving massive investments. Five large listed companies - Amazon.com, Inc. (AMZN.US), Alphabet Inc. Class C (GOOGL.US), Microsoft Corporation (MSFT.US), Oracle Corporation (ORCL.US), and Meta Platforms (META.US) - are currently making large investments in AI. However, their stock prices have decreased by an average of about 6% so far this year. The huge capital expenditure requirements are forcing tech giants to raise large amounts of funds through debt and stock issuance. For example, SpaceX (SPCX.US) announced the issuance of $20 billion in bonds, joining this AI arms race and continuing to increase AI investments along with Amazon.com, Inc., Alphabet Inc. Class C, Microsoft Corporation, Oracle Corporation, and Meta. In contrast, semiconductor sector-related ETFs - including VanEck Semiconductor ETF (SMH), iShares Semiconductor ETF (SOXX), SPDR S&P Semiconductor ETF (XSD), Direxion Daily Semiconductor Bull 3X Shares ETF (SOXL), etc. - have seen overall gains. This further confirms Dan Niles' views. He expressed his concerns and pointed out that enterprise AI strategies are undergoing a significant shift. Just a few months ago, companies were focused on "Token maximization," encouraging employees to generate as many AI Tokens as possible. Now, the focus has shifted to "Token minimization" as companies gradually realize that they cannot exhaust the entire year's AI budget in a few months. He said: "Like Uber Technologies, Inc., you can't spend the entire AI budget in four months without any issues when announcing earnings and providing guidance." Looking ahead, Dan Niles warned that companies are expected to report strong results for the June quarter due to their previous push for "Token maximization." However, as corporations begin to route more requests to lower-cost open-source models - some of which cost only one-eighth of high-end models - guidance for the September quarter may come under pressure. He said: "My question is if you are shifting tasks to cheaper models, what will your September guidance look like?" "This is where I think these companies may hit a 'speed bump' when announcing earnings and issuing future guidance." Regarding his investment portfolio strategy, Dan Niles stated that he is gradually reducing positions in large-scale cloud computing companies and also starting to trim some semiconductor holdings - although the semiconductor sector has shown strong performance. He pointed out that the semiconductor index has doubled, but emphasized that as AI-related investment logic continues to evolve, investors need to become "more discerning." Dan Niles' views reflect a reality where as more investors begin to question whether the massive capital expenditure of large-scale cloud service providers on AI infrastructure expansion is reasonable, the threshold for AI trading is rising. Merely talking about future potential is not enough, as the market now demands tangible orders, cash flow, and profit realization. This means that investors will turn to sectors benefiting from the AI capital expenditure wave, including the semiconductor industry. Columbia Threadneedle Investments, a fund management company, also holds a similar view. The organization believes that as investment in AI infrastructure continues to accelerate, the upward trend in technology stocks is likely to continue for at least the next few quarters, with expectations for revenue and profit of related companies still being continually raised. However, as the AI competition enters a new stage, changes are occurring in the flow of funds and competitive landscape within the market, with investors increasingly focusing on which companies can truly benefit from the AI capital expenditure wave. Tiffany Wade, Senior Portfolio Manager at Columbia Threadneedle Investments, stated in an interview on Monday that companies related to AI infrastructure are still in a strong growth cycle, with market expectations for their revenue and profit prospects continually improving. "Technology companies related to AI infrastructure spending have very promising prospects, with their revenue and profit forecasts continuously being revised upwards," Wade said. Despite a slight cooling of leading U.S. technology stocks on Monday, which noticeably pressured Asian technology stocks on Tuesday, Tiffany Wade believes this is more of a short-term volatility and will not change the long-term trend of the expanding AI investment cycle. She noted that recent industry dynamics show that AI infrastructure construction is still accelerating. For instance, Micron Technology, Inc. (MU.US) reached an AI infrastructure cooperation agreement with Anthropic, and multiple large data center projects in Texas, USA, continue to expand, reflecting a strong growth momentum in AI computing power demand. Tiffany Wade expects this trend to continue for at least the next few years. Some analysts pointed out that although global technology stocks have shown relative weakness recently, this adjustment may be more like short-term market volatility due to political events like the GEO Group Inc and changes in investors' confidence in the sustainability of the AI market. In the medium term, factors like the expansion of capital spending by U.S. big factories, whether AI companies can catch up with valuation for profitability, and whether the Asian semiconductor chain can maintain order and price expectations will be the market's focal points.