The AI market rally is approaching a significant test. Dan Niles, a well-known semiconductor analyst from the internet bubble era and founder of Niles Investment Management, has cautioned investors that the AI trade is about to hit a "speed bump." He argues that smart money should flow to where AI capital is actually being spent—the semiconductor industry—rather than to the hyperscale cloud companies that are making massive investments.
Major public companies like Amazon.com (AMZN.US), Alphabet (GOOGL.US), Microsoft (MSFT.US), Oracle (ORCL.US), and Meta Platforms, Inc. (META.US) are currently investing heavily in AI. However, their stock prices have declined by an average of approximately 6% year-to-date. The enormous demand for capital expenditure is compelling these tech giants to raise substantial funds through debt and equity issuance. For instance, SpaceX has announced a $20 billion bond offering to join the AI arms race, continuing to ramp up investments alongside Amazon, Google, Microsoft, Oracle, and Meta.
In contrast, semiconductor-related ETFs, including VanEck Semiconductor ETF (SMH), iShares Semiconductor ETF (SOXX), SPDR S&P Semiconductor ETF (XSD), and Direxion Daily Semiconductor Bull 3X Shares (SOXL), have seen overall gains. This trend further supports Dan Niles's perspective. He elaborated on his concerns, highlighting a significant shift in corporate AI strategy. Just a few months ago, companies were focused on "token maximization," encouraging employees to generate as many AI tokens as possible. Now, the focus of discussion has shifted to "token minimization," as businesses realize they cannot exhaust their annual AI budgets within a few months.
He stated, "Just like with Uber, you can't spend your entire AI budget in four months without encountering problems when reporting earnings and providing guidance." Looking ahead, Niles warned that due to the previous corporate push for "token maximization," June quarter results are still expected to be quite strong. However, as companies begin routing more requests to cheaper open-source models—some costing as little as one-eighth of high-end models—September quarter guidance could come under pressure.
"My question is, if you are shifting tasks to cheaper models, what will your September guidance look like?" he said. "This is precisely where I think these companies might hit a 'speed bump' when reporting earnings and issuing future guidance." Regarding his portfolio strategy, Niles indicated he is gradually reducing exposure to hyperscale cloud companies and has also started trimming some semiconductor holdings, despite the sector's strong performance.
He noted that "the semiconductor index has doubled," but emphasized that as the investment thesis around AI evolves, investors need to become "more selective." Niles's view reflects a growing reality: the bar for the AI trade is rising. As more investors question the rationale behind the massive capital expenditures by hyperscale cloud providers on AI infrastructure expansion, simply touting long-term potential is no longer sufficient. The market is beginning to demand visible orders, cash flow, and profit realization.
This implies a shift in investor focus towards areas that benefit from the AI capital expenditure wave, including the semiconductor industry. Asset manager Columbia Threadneedle Investments shares a similar view. The firm believes the tech stock rally, driven by accelerating AI infrastructure investment, is likely to continue for at least the next few quarters, with revenue and earnings expectations for related companies still being revised upwards.
Shifting Investment Landscape and Focus
However, as the AI race enters a new phase, capital flows and the competitive landscape within the market are changing. Investors are paying closer attention to which companies can genuinely benefit from the AI capex wave. Tiffany Wade, a senior portfolio manager at Columbia Threadneedle Investments, stated in an interview that companies related to AI infrastructure are still in a strong growth cycle, with market expectations for their revenue and profit prospects continuously improving.
"The outlook for tech companies tied to AI infrastructure spending is very significant, and their revenue and earnings forecasts are still being revised upward," she said. Although U.S. tech leaders cooled on Monday, dragging down Asian tech stocks significantly on Tuesday, Wade views this more as short-term volatility that does not alter the long-term trend of an expanding AI investment cycle.
She pointed out that recent industry developments indicate AI infrastructure build-out is still accelerating. For example, Micron Technology's AI infrastructure partnership with Anthropic and the continued expansion of multiple large data center projects in Texas, USA, reflect that demand for AI computing power remains robust. Wade expects this trend to persist for at least several more years.
Market Outlook and Key Factors
Some analysts note that although global tech stocks have shown relative weakness recently, this correction may resemble short-term market volatility driven by geopolitical events and shifts in investor confidence regarding the sustainability of the AI rally. From a medium-term perspective, key market focus areas will include whether capital expenditure by major U.S. firms continues to expand, whether AI company profits can catch up to valuations, and whether the Asian semiconductor supply chain can maintain order and pricing expectations.