Author: Chasing Wind Trading Platform, Wall Street Watch

The AI market is not simply a repeat of the 1999-2000 bubble. Goldman Sachs believes the more critical issue now is that earnings and capital expenditure are still being revised upwards, but market prices have already factored in a large amount of optimistic expectations, and investor sensitivity to narrative changes is increasing.

According to Chasing Wind Trading Platform, Goldman Sachs predicted in its June 22 research report that the AI investment boom may continue, and the recent market expectations for its scale may even need to be adjusted upwards. However, the report also pointed out that a lot of value has already been reflected in advance, and the market will be more vulnerable to any challenges to the optimistic narrative surrounding AI.

The main risk in AI trading is no longer just "valuation bubbles". Forward-looking price-to-earnings ratios have not been significantly out of control, because earnings expectations have been revised upwards simultaneously. What really needs to be tested is whether current strong earnings can continue after the capital expenditure cycle peaks.

For investors, strong earnings may continue to overshadow valuation concerns before the peak of the AI investment cycle occurs. However, as the incremental market value increasingly relies on optimistic assumptions, stock volatility may rise further, and the value of downside protection is also increasing.

AI is not 1999, but the market has already run ahead of the macroeconomic situation

Goldman Sachs's core judgment is that today's AI cycle is not characterized by extreme valuation expansion, macroeconomic overheating, and financing imbalances as seen in 1999-2000.

Currently, the fundamentals have not deteriorated significantly; in fact, they are still strengthening. AI-related companies are experiencing strong earnings, and capital expenditure plans continue to be revised upwards, giving the market reason to continue buying related assets. Compared with the late 1990s, forward-looking valuations have not experienced an equivalent degree of loss of control.

However, this does not mean that risks are lower. The market capitalization growth of AI-related companies has significantly exceeded benchmark macroeconomic profit estimates. To explain current prices, one must assume that AI winners can maintain productivity premiums above normal levels over the long term.

In other words, the core bet that the current market is making is not "valuation can expand infinitely" but "super high profits can persist".

The true similarity to the 90s is the intensity of investment; other bubble signals have not yet emerged

In the late 1990s tech bubble, there were four typical signals: investment maintained at unusually high levels, macro profits declined, corporate financing demand and leverage rose rapidly, and the current account deficit widened.

Currently, the only significant signal that has emerged is the first, which is the acceleration of AI capital expenditure. The research report states that the proportion of technology investment in GDP has already surpassed the peaks of the 1990s and is rising faster. The expectations for 2026 capital expenditure from super-large cloud providers have increased by nearly 80% compared to six months ago. Based on current trajectories, AI-related investment could approach or even exceed the peak levels of the tech investment boom in the 1990s in the coming years.

However, this capital expenditure cycle differs from that of the past. Firstly, its duration has not yet reached the length seen in the late 1990s, and secondly, its scope is not as broad as back then. The technology investment of the 1990s resembled an expansion across the entire economy, while today’s AI capital expenditure is more concentrated among super-large cloud providers, semiconductors, and related infrastructure chains.

The most critical contrast at the macro level lies in profits.

In the late 1990s, corporate profit margins peaked and fell after 1997, with rising wages and unit labor costs eroding profits. The current situation is different; corporate profits as a share of GDP remain close to high levels, and productivity growth has not been fully offset by the acceleration in wages as it was back then.

The corporate financing side has also not replicated the path seen then. Super-large cloud providers have seen a significant decline in free cash flow, and the proportion of capital expenditure relative to operating cash flow has risen sharply. However, across the entire corporate sector, the gap between savings and investment has not deteriorated significantly, as profit growth has largely mitigated the rising investment rate.

External imbalances are also different. In the late 1990s, the U.S. current account deficit widened; currently, the current account deficit is actually narrowing. At least from the perspective of macroeconomic imbalance, the current AI cycle has not displayed the typical cracks seen at the end of the bubble in those years.

$27 trillion market value increase, exceeding the benchmark macro account

Changes at the market level are more radical.

Since the end of November 2022, the value increment of AI-related companies has been approximately $27 trillion, surpassing the level of about $19 trillion in November 2025. Meanwhile, traditional valuations in the U.S. stock market remain at historical highs, with the Shiller cyclically adjusted price-to-earnings ratio only having been higher at the end of 1999 and in 2000.

However, this increase differs from that of 1999 in a key way: earnings expectations are also being revised upwards rapidly. Due to rising EPS expectations, even if stock prices continue to rise, forward-looking price-to-earning ratios have not correspondingly increased this year. Recent gains have been driven more by earnings rather than mere valuation expansion.

The issue lies in the macro account not providing support of equivalent scale. Benchmark estimates show that new capital income from AI productivity improvements has a present value of approximately $9 trillion for the U.S. economy. Even under a more conservative market lens, looking only at "pure AI" companies, the related value increment is about $14 trillion; if we add a 25% increase from other AI-related companies, the total scale is about $17 trillion, still exceeding the benchmark estimates.

To support current prices, one must bet that winners will capture higher profits long-term

Current market prices are not entirely inexplicable, but they require more optimistic assumptions.

These assumptions include: AI adoption is faster, the productivity boost from AI is higher, capital claims a larger share of economic income, or American companies can obtain more of the global AI revenue.

One optimistic trajectory given in the research report is that U.S. companies capture 50% of global related revenues, capital income shares significantly exceed the economic average, AI adoption is faster, and discount rates are lower. Only if multiple conditions are met simultaneously is it easier for potential value to cover the current market value increase.

The most persuasive optimistic narrative is that AI-related companies can capture a higher share of the productivity premiums in the long run. So far, this narrative has indeed been supported by profits. Profits from semiconductors, cloud providers, and infrastructure beneficiaries are robust, with high margins, and it is these earnings that support the market.

But this is also its weakness. During the initial phase of productivity acceleration, profit shares typically rise; over time, competition, investment expansion, and a new round of innovation may erode excess returns. The AI industry has a high concentration, and the technological characteristics may favor capital owners; however, how long the barriers maintained by current winners will last remains uncertain.

The greatest risk shifts from "valuation bubbles" to "profit bubbles"

The AI investment boom itself is generating substantial profits. Companies that sell chips, provide computing power, and build data centers directly benefit from rising capital expenditure. As long as peak investment has not yet approached, earnings revisions may continue to overshadow valuation concerns.

However, if the market directly extrapolates strong earnings for the next two to three years into a more distant future, risks will increase. Capital expenditure cannot grow at the current strength indefinitely. Once the investment cycle peaks, it may become more challenging to predict the earnings curve for the companies that currently benefit the most directly.

This is also why "forward-looking price-to-earnings ratios are not expensive" does not necessarily mean they are cheap. Cyclical industries and commodity companies often appear cheap at cyclical peaks because the earnings denominator is too high. Whether the AI infrastructure chain will encounter similar issues depends on how long the intensity of investment can be sustained, how quickly AI profits are realized, and whether technological innovations can reduce dependence on high-intensity capital expenditure.

AI may be masking the weakness of the non-AI economy

Compared to the 1990s, there is another significant difference in the current macro backdrop.

At the end of the 1990s, U.S. domestic demand was very strong, with actual domestic demand annualized growth rates close to 6% in the last two years; consumption, residential investment, and non-technology investment were all robust. Capital inflows resulting from the Asian and emerging market crises, a strong dollar, and global commodity price deflation masked internal overheating in the U.S., allowing the cycle to extend longer.

Currently, the situation is the opposite. The U.S. economy outside of AI is not as strong. Non-technology investment is weak, and consumption growth is far from the late 1990s, with actual disposable income annualized growth rates in the past two years around 1%, compared to 5%-6% in the late 1990s.

This implies that the AI boom may not be fueling a fully overheated economy but is compensating for weakness in areas outside of AI. Consequently, the extreme bubbles and typical imbalances seen before the recession in 2001, such as those in 1999-2000, may be less likely to occur; however, if the AI narrative falters, the non-AI sectors may not provide sufficient support.

Volatility shifts, portfolios require more downside protection

Market structure has already changed.

Credit spreads remain tight, differing from the gradual increase in credit pressures seen from 1998-2000. However, stock volatility has started to rise more noticeably. In recent months, implied volatility of individual stocks has increased, the skew of U.S. single-stock options has shifted down, and the demand for call options relative to put options has risen.

At the same time, implied correlations have dropped to very low levels, compressing index volatility, but long-term index volatility is also slowly rising. Gains have also become more concentrated. Broad index performance is still milder than the late 1990s, but the semiconductor index has nearly approached the performance of the NASDAQ in its later stage over the past few years. In April and May, the consecutive two-month gains of the NASDAQ, South Korea, Taiwan, SOX semiconductor index, and the basket of unprofitable tech stocks all reached multi-year highs.

As long as the peak of the investment cycle has not yet occurred, strong earnings may continue to dominate the market. However, as prices increasingly depend on optimistic assumptions, the value of downside protection rises. In terms of strategy, it resembles staying in the trade while employing put protection or replacing some spot exposure with call options to control drawdowns.

There is also a reverse risk on the interest rate side: if the peak of AI investment passes and the vulnerability of the non-AI economy is exposed, the probability of significant interest rate declines may exceed usual levels.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。