As AI capital expenditures surge and valuations become increasingly expensive, Goldman Sachs reminds the market: the real risk often appears at the moment when &#34;growth begins to slow.&#34;
On February 24, Goldman Sachs Global Investment Research stated in its strategy report &#34;The Broadening and Narrowing of AI Trades&#34; that recent volatility in AI trades has significantly increased, driven by two opposing forces in the market: on one side, the capital expenditures of tech giants continue to exceed expectations, while on the other side, investor concerns about &#34;AI disrupting traditional industry profit pools&#34; are rapidly escalating.
Driven by strong capital expenditure guidance, storage chip concept stocks have surged an average of 55% so far this year; meanwhile, software stocks have plummeted 24% due to panic over the &#34;AI disruption theory.&#34; The same &#34;AI theme&#34; is showing almost opposite market trends at different stages.
Goldman Sachs categorizes this extreme volatility in AI trades into four stages, with the stock price trends of these four stages currently being completely opposite:
<blockquote>
<ul>
<li>Stage One (Leading computing power, such as NVIDIA): Facing questions of &#34;excessive profitability,&#34; with a recent disconnect between significantly raised profit expectations and stagnant stock prices.</li>
<li>Stage Two (AI infrastructure, such as storage, devices, servers, etc.): Driven by strong capital expenditure guidance from tech giants, these stocks have recently surged, with storage stocks up 55% this year.</li>
<li>Stage Three (AI application empowerment, such as software services, etc.): Due to extreme market concerns about their traditional business models being disrupted by AI, these stocks have recently faced panic selling, with software stocks down 24% this year.</li>
<li>Stage Four (AI productivity enhancement, non-tech industries): Due to unclear actual financial returns, stock prices have remained stagnant recently.</li>
</ul>
</blockquote>
In the face of this extreme differentiation, the report indicates that both the currently surging &#34;infrastructure winners&#34; and the plummeting &#34;application losers&#34; harbor their own risks.
<h2>Capital expenditure growth rate nearing its peak, &#34;valuation killing&#34; risk for AI infrastructure approaching</h2>
The market's first task is to digest the &#34;further upward adjustment&#34; of capital expenditure expectations.
According to consensus expectations compiled by Goldman Sachs, AI capital expenditures by tech giants are expected to reach $667 billion by 2026. This figure is $127 billion higher than at the start of the fourth quarter earnings season, with a year-on-year growth rate of 62%.
On the other side of the significant upward revision of capital expenditures is the squeeze on free cash flow.
The report emphasizes: &#34;The capital expenditures of super cloud vendors are heading towards exceeding 90% of operating cash flow this year, a proportion even higher than during the internet bubble.&#34; In more specific calculations, Goldman Sachs points out that capital expenditures are expected to account for 92% of the operating cash flow of tech giants by 2026.
To fill the massive funding gap, these giants are forced to significantly reduce shareholder returns. In 2025, overall stock buybacks by these giants were cut by 15%; the proportion of cash flow used for buybacks plummeted from 43% at the beginning of 2023 to the current 16%. At the same time, companies like Oracle and Google have begun to frequently reach out to the bond market.
Goldman Sachs expects that there is still room for upward revision of the absolute value of capital expenditures this year. As Oracle and Microsoft’s fiscal years end in May/June, the upcoming second-quarter earnings season may serve as a catalyst for further upward adjustments in expenditure expectations
However, Goldman Sachs warns that the core risk lies not in the &#34;absolute value,&#34; but in the &#34;growth rate.&#34;
<blockquote>
“We expect that consensus estimates for capital expenditures of super cloud vendors still have moderate upside potential, but we anticipate that the growth rate of capital expenditures will peak later this year.”
</blockquote>
This slowdown in growth will become the &#34;Achilles' heel&#34; of AI infrastructure stocks.
<h2>Risks in the Second Half of &#34;AI Infrastructure&#34;: Slowing Expenditure and the &#34;Over-Earning&#34; Trap</h2>
Goldman Sachs emphasizes: “Once the growth rate of capital expenditures slows, the revenue growth and valuations of some AI infrastructure stocks will appear extremely fragile.”
The logic is straightforward: Orders, revenue, and profits along the infrastructure chain are often highly sensitive to the growth rate of capital expenditures; when the market shifts from &#34;accelerating every quarter&#34; to &#34;still growing but no longer accelerating,&#34; the most vulnerable part of the valuation is often the &#34;growth premium.&#34;
Goldman Sachs bluntly states that many AI infrastructure-related sectors have experienced significant valuation multiple expansions over the past few years, and historical experience shows that investors typically assign lower valuation multiples to companies with &#34;slowing growth.&#34;
This is also the core meaning of the so-called &#34;valuation kill&#34; in the report: Even if profits are still growing, as long as the market begins to worry about the sustainability of growth, multiple contractions may offset the price support brought by profit upgrades.
In the sub-sectors listed in the report, valuations in areas such as manufacturing equipment, servers and networks, foundry and IDM, power and utilities are generally higher than the average of the past five years.
Goldman Sachs believes that the current &#34;latest bottleneck&#34; within infrastructure is concentrated in the memory segment.
The report states that major memory stocks (such as Micron, Western Digital, SK Hynix, Samsung) have averaged an increase of about 145% since the beginning of Q4 2025, with an average increase of about 55% this year. Goldman Sachs believes that strong demand and price increases account for most of the sources of this rise.
They also point out that memory stocks have an average forward P/E of about 12 times, which is not only lower than the market but also below their own five-year average, appearing not &#34;expensive&#34; on the surface.
However, Goldman Sachs immediately uses NVIDIA as a warning: When the market begins to worry that companies are in an &#34;over-earning&#34; state, stock prices may no longer follow profit upgrades.
From the end of 2022 to mid-last year, NVIDIA's stock price and earnings grew synchronously by 12 times, with valuation multiples remaining basically unchanged. But in the recent phase, the logic has changed.
Goldman Sachs points out: “In the past five months, despite a significant 37% upward revision in NVIDIA's forward earnings expectations, its stock price has remained basically flat.”
Goldman Sachs summarizes this phenomenon as the market psychology of &#34;over-earning&#34;: when a company performs too strongly at the peak of the cycle, it tends to attract concerns about intensified competition and the sustainability of demand, ultimately manifesting as &#34;profits continue to be strong, but valuations contract.&#34;
From a trading perspective, this means: Even if the performance of the infrastructure chain continues to deliver in the short term, investors will begin to be more selective about the &#34;second derivative of growth&#34; and whether multiples can still expand.
<h2>Tech Giants Continue to Diverge in the Short Term: Focus is Not on Capital Expenditure, but on &#34;Returns&#34;</h2>
Goldman Sachs predicts that the divergence in returns among tech giants will continue in the short term.
As capital expenditure growth stabilizes in the first half of 2026, market attention will shift to whether &#34;AI investments yield returns.&#34;
The report provides a set of intuitive comparisons: the free cash flow yield of tech giants is about 1%, at a historical low; while the rest of the S&amp;P 500 companies is around 4%.
When free cash flow weakens and conversion rates decline, funds will naturally seek alternative options. Goldman Sachs bluntly states, &#34;Investors are increasingly reallocating their funds elsewhere.&#34;
<h2>AI Application Layer: A &#34;Very Thin Line&#34; Dividing Winners and Losers</h2>
If the contradiction at the infrastructure level is &#34;how fast can capital expenditure grow,&#34; the contradiction at the application layer is &#34;who will be disrupted and who can capture new revenue.&#34;
Goldman Sachs judges that the spread of AI transactions to the application layer is a natural path of technological development: once the infrastructure is in place, value creation will shift from &#34;selling shovels&#34; to &#34;transforming business models,&#34; and recover initial investments by reshaping profit pools.
However, this also makes stock market outcomes more &#34;micro-focused.&#34; Goldman Sachs emphasizes that future assessments will need to rely more on company-level judgments, such as competitive positioning, entry barriers, and pricing power.
A statement in the report highlights the core uncertainty at the application layer:
<blockquote>
&#34;In an uncertain final competitive landscape, the line between a company being seen as an 'AI revenue winner' and facing 'disruption' concerns is very thin.&#34;
</blockquote>
One direct result is that investors are currently not giving many listed companies high valuations for &#34;AI bringing in new revenue.&#34;
Goldman Sachs states, &#34;Contrary to our expectations, investors have priced almost no upside for listed companies' AI revenue growth; instead, AI applications from private companies have received the most attention.&#34;
The report lists the product progress of several private companies: Anthropic launched the Claude Cowork tool (including legal, human resources, and business service plugins); Insurify launched a comparison application within ChatGPT; Altruist introduced a tool for creating personalized tax strategies for wealth management clients.
These cases reinforce a concern in the public market: even if AI generates new demand, the new revenue may not necessarily belong to listed companies.
<h2>Why &#34;Losers&#34; Find It Hard to Turn Around in the Short Term: Disruption Concerns Are Difficult to Disprove with &#34;Short-Term Performance&#34;</h2>
On the other side of the application layer is the destructive power of disruption narratives on valuations.
Goldman Sachs points out that market focus has shifted to &#34;AI disruption risks&#34; in recent weeks.
The report states that software stocks have fallen about 23% in the past six weeks, and &#34;despite short-term earnings remaining resilient, investors are increasingly questioning the industry's long-term growth prospects.&#34;
Goldman Sachs provides a very clear judgment here: &#34;Concerns about AI disruption are difficult to disprove in the short term.&#34;
They further point out: for those companies already labeled by the market as &#34;potentially disrupted by AI,&#34; stabilizing stock prices hinges on stabilizing earnings first; however, &#34;this uncertainty of disruption is unlikely to be resolved in the short term.&#34;
Goldman Sachs has specified the conditions under which &#34;losers at the application layer will find it difficult to turn around in the short term&#34;: &#34;Investors need either multiple quarters of evidence proving business resilience or a significant valuation decline of these stocks relative to the market before they will re-engage on a large scale.&#34;
This is also the awkward situation for sectors like software currently: short-term financial reports may be acceptable, but the market is trading on &#34;whether the long-term profit pool will be redistributed.&#34;
<h2>Goldman Sachs Quantifies Disruption Risk with Two Clues: Exposure to AI and Asset Intensity</h2>
In observing &#34;who is more likely to be disrupted,&#34; Goldman Sachs provides two vectors (and emphasizes that there are other dimensions such as regulatory barriers and market power).
First, the exposure of labor to AI automation.
Goldman Sachs states that concerns about the replacement of white-collar jobs have risen recently.
They collaborated with economists to estimate the proportion of wage expenditures at various companies that are exposed to AI automation, and observed this in conjunction with the ratio of &#34;labor costs/income.&#34;
Goldman Sachs warns that this indicator is a &#34;double-edged sword&#34;: AI can both enhance efficiency and replace jobs.
However, on the trading front, the market has rewarded industries with &#34;low exposure&#34; over the past six months and punished those with &#34;high exposure.&#34;
Second, tangible asset intensity.
Goldman Sachs measures asset intensity using &#34;(assets - cash - intangible assets)/income&#34; and constructs an industry-neutral, equal-weight basket.
They observed that companies with heavier assets have significantly outperformed those with lighter assets recently, and the extent of this outperformance &#34;exceeds what the macro environment can typically explain.&#34;
Similarly, companies producing goods have outperformed service-oriented companies.
For investors, these two clues do not convey that &#34;heavier assets are better,&#34; but rather that the market is using them as &#34;substitute indicators for moats/barriers to entry&#34; to combat uncertainties at the application layer.
<h2>Three Catalysts: Goldman Sachs Pins the &#34;Turning Point&#34; on the Second Half of 2026</h2>
Goldman Sachs believes that three catalysts are needed for tech giants to regain market leadership.
Their baseline judgment is that these catalysts &#34;are more likely to emerge in the second half of 2026.&#34;
First, AI revenue must accelerate. The market reaction during earnings season has already proven that as long as revenue growth exceeds expectations (such as Meta's 10% surge), investors will regain confidence in the return on AI investments.
Second, the visibility of free cash flow (FCF) bottoming out due to slowing capital expenditure growth. Goldman Sachs believes that once cash flow shows bottoming signals, the market may reprice based on earnings rather than cash flow, thereby reducing valuation volatility.
Goldman Sachs explains: &#34;The slowdown in capital expenditure growth will give investors hope for a rebound in free cash flow. This will prompt investors to reprice these giants based on profitability.&#34; Currently, the giants' forward P/E ratio of 24 is only at the 14th percentile of the past decade, making valuations highly attractive.
Finally, the fading of macro tailwinds. Goldman Sachs economists expect the cyclical acceleration of the U.S. economy to peak around mid-year and then decline in the second half. As macroeconomic dividends fade, funds will inevitably flow back to these tech giants with very high long-term certainty
The above wonderful content comes from Chasing Wind Trading Platform.
For more detailed interpretations, including real-time analysis and frontline research, please join the【 Chasing Wind Trading Platform ▪ Annual Membership】.

NVDA

Goldman Sachs warns that the growth rate of AI capital expenditures is expected to slow in the second half of the year. This turning point will directly threaten AI infrastructure stocks that heavily rely on capital expenditures, causing their valuation premiums to collapse. For example, NVIDIA is currently experiencing the phenomenon of "profits soaring, but stock prices not rising." On the other hand, software and other AI application companies deeply mired in "disruption" panic cannot dispel the market's long-term concerns solely based on short-term financial reports, making it extremely difficult to turn things around in the short term

- Goldman Sachs warns that true risks in AI investments often emerge when growth begins to slow, as capital expenditures surge amidst rising valuations.  
- Recent fluctuations in AI trades are driven by strong spending from tech giants, which has led to significant gains in storage stocks while software stocks have plummeted due to fears of AI disrupting traditional business models.  
- As capital expenditure growth is expected to peak, risks associated with valuation compression in AI infrastructure stocks loom, with investor concerns about sustainability overshadowing ongoing earnings growth.

Goldman Sachs on AI Trading: Risks of "AI Infrastructure" in the Second Half, "Losers" of "AI Applications" Hard to Turn Around in the Short Term