Agent AI consumption surged 100 times, why did the market repeat the 1998 crisis?

According to Woofun AI, the artificial intelligence market is facing the most severe structural test since the 1997-1998 internet bubble burst. Although Dan Niles is convinced that the explosive growth of agent-based artificial intelligence (Agentic AI) means the industry is far from peaking, he recently continued to recommend investors increase their cash reserves to deal with systemic risks.

This cautious attitude is particularly prominent during the critical window period when Microsoft (MSFT.US), Alphabet (GOOGL.US), and Amazon (AMZN.US) are about to release financial reports. Last week, the PHLX Semiconductor Index (SOX) plummeted 3.2%, while the S&P 500 index fell slightly by 0.5%. The sharp retracement of chip stocks revealed a sharp divergence between market sentiment and fundamentals. Niles pointed out that proxy AI consumes 10 to 100 times the number of tokens in operation. This technology evolution logic supports long-term bull market expectations, but short-term valuation revisions are strikingly similar to historical cycles, and investors must find a balance between the long-term dividends of technological change and short-term market turmoil.

The underlying reason lies in the double deterioration of the financing environment and growth momentum. According to data compiled by Woofun AI, financing costs in debt and equity markets are rising at the same time, directly impacting the valuation system of companies with AI as their core business. Take CoreWeave (CRWV.US) as an example. In June, the company issued an unsecured bond with a face value of $1.25 billion and maturing in 2032, with an initial coupon interest rate of 9.625%;

However, as of last week, its bond yield had soared to about 15%, while the stock price had fallen 36% cumulatively since the beginning of October. As CoreWeave (CRWV.US) provides critical computing services to Microsoft (MSFT.US) and Meta (META.US), its financing difficulties have spread deep into the AI supply chain of large tech companies.

At the same time, signs of slowing growth are also evident. Anthropic's quarterly revenue growth rate plummeted from 145% in the second quarter to 45% in the third quarter. This cliff-style decline in growth further heightened market concerns about the pace of AI commercialization.

The more critical variables are political and regulatory risks. The agreement reached between the two parties in the US on suspending data center construction may be transformed into stricter implementation standards after the midterm elections, thereby suppressing cloud service business expectations for the December quarter. Australia's NextDC was forced to lower its performance guidelines due to community opposition, and its stock price plummeted 39.5% in a single day on October 21. This case vividly illustrates the instantaneous impact of geographical and community resistance on asset prices. On the mainland of the US, Amazon (AMZN.US) has abandoned plans to build four data centers in Loudon County, Virginia, with a total investment of 427 million US dollars. From April to June alone, $68 billion worth of data center projects in the US were suspended or postponed due to opposition. This massive capital expenditure standstill is reshaping the industry landscape.

Microsoft (MSFT.US) and Alphabet (GOOGL.US) are scheduled to release financial reports after the close of trading on October 28, just a few hours after the announcement of the Federal Reserve's interest rate decision; Amazon (AMZN.US) will disclose results on October 29. Niles stressed that the stock price trend of these tech giants will be highly dependent on their statements on capital expenditure, operating costs, cash flow conditions, and future financing needs. In particular, he warned that even with strong financial performance in the September quarter, the suspension of data center construction could have a substantial negative impact on cloud service business development expectations for the December quarter.

This lack of expectations may be a key catalyst for triggering a new round of market adjustments. Especially in a context where the market is already fatigued by high growth narratives, any signal that infrastructure expansion is blocked may be amplified and interpreted.

Divergent market opinions further underscore the complexity of the current situation. Former BitMEX CEO Arthur Hayes believes that excessive expansion in the field of artificial intelligence will eventually cause a crisis, while Tom Lee of Fundstrat argues that the AI-related sell-off is nearing its end. Financial data and financing market trends to be released this week will be a key litmus test for these two opposing views. Following the collapse of the Internet bubble in 1998, the technology industry once again faced a fierce collision between technological beliefs and capital reality. Investors need to pay close attention to marginal changes in capital expenditure efficiency and policy regulation.