US tech stocks polarization: capital abandons "shovel sellers" to chase "gold diggers", AI market enters realization period

In early August 2026, the US tech sector staged a "song of ice and fire" that left investors amazed. On one side, Google, Amazon, and Microsoft, the three major cloud giants, successively refreshed historical records in stock prices, with Amazon's market value historically breaking through the $3 trillion mark, and Google's total market value surpassing Apple to rank second globally; on the other side, the previously brilliant semiconductor sector collectively faced pressure, with the Philadelphia Semiconductor Index plummeting over 20% in July. Under the same AI wave, why are tech stocks taking two completely opposite curves? The answer may point to a deeper signal—the AI market is entering a new stage of "verifying cash" from "storytelling".

Ice and fire: cloud giants hit new highs, semiconductors under pressure

On the morning of August 3, US time, the Wind US Tech Seven Giants Index (MAGS.WI) rose 3.03% in a single day, with Meta rising over 6%, Microsoft rising nearly 5%, Amazon and Google rising 4.58% and 4.44% respectively, Tesla rising 3.49%, and NVIDIA rising nearly 3%, with Apple falling 1.78% against the trend. Thanks to this strong performance, Google's total market value surpassed Apple to climb to second place globally, while Amazon's market value stood on the $3 trillion step for the first time.

In sharp contrast to the high-spirited advance of cloud giants, the US semiconductor sector has been under overall pressure recently. The Philadelphia Semiconductor Index experienced a violent adjustment in July, once plummeting over 20%, with market doubts about the return rate of AI capital expenditure集中宣泄. However, on August 4, the index surged 6.55% in a single day, with ARM rising over 17%, and storage chip stocks rebounding significantly, showing that the kinetic energy of oversold recovery cannot be underestimated.

This differentiation is not a one-day affair. Xiangsong Capital Executive Director Shen Meng pointed out that the US market is currently in a sector rotation period of the AI industry chain: the semiconductor sector has risen too much and valuations are high, while the cloud computing sector is in a relative valuation lowland, with capital rebalancing between the two. Nankai University Finance Professor Tian Lihui described the current situation as a "historical gear shift"—global tech investment logic is undergoing profound changes.

Earnings set the tone: AI from "story" to "revenue"

The direct catalyst for this sector rotation is the better-than-expected earnings reports successively released by the three major cloud giants. On July 23, Google's parent company Alphabet released its second quarter 2026 earnings report, with revenue of $119.796 billion in the quarter, a 24% year-on-year increase, and net profit of $112.1 billion, a surge of 298% year-on-year. On July 29, Microsoft released its 2026 fiscal year fourth quarter performance, with revenue of $90.01 billion, a 18% year-on-year increase, and net profit of $35.8 billion, a 31% year-on-year increase. On July 31, Amazon released its second quarter report, with revenue of $200.606 billion, a 20% year-on-year increase, and net profit of $62.647 billion, a surge of 245% year-on-year.

The three earnings reports have a common highlight: cloud business growth has fully recovered. Amazon's AWS-related revenue grew 37%, Microsoft's cloud business revenue grew 27%, and Alphabet's operating profit margin rose to 34%. This means that the past two years of trillions of dollars in AI infrastructure investment are beginning to be converted into real revenue through cloud computing, enterprise software, and advertising efficiency improvements, and the AI narrative is obtaining endorsement from business data.

A more convincing indicator is the order backlog of cloud providers. According to market institutions, the order backlog of large-scale cloud service providers' cloud business grew by over 150% year-on-year, reaching about $1.7 trillion, far higher than the同期 about 80% capital expenditure growth. Order growth outpacing capital expenditure indicates that new computing power is not falling into oversupply but is quickly absorbed by strong demand—this is the core basis for Wall Street to reassess AI infrastructure return capability.

Profit quality interrogation: "paper wealth" emerges

However, behind the bright report numbers, there is also a detail worth alerting: some profits do not come from main business but from the book value appreciation of equity investments in AI startups held by tech giants.

FactSet data shows that the S&P 500 index components' second quarter profit grew by 47.4% year-on-year, which seems quite amazing on the surface; but if Alphabet and Amazon are excluded, the growth rate drops sharply to 28.8%. Goldman Sachs further calculated that the S&P 500's overall EPS soared 45% year-on-year, of which about 19 percentage points came from Alphabet and Amazon's unrealized gains from equity investments, and after excluding them, the actual growth rate was about 26%. Amazon confirmed $53.4 billion in non-operating pre-tax income, mainly from its investment in Anthropic; Alphabet recorded $98 billion in "other income", mainly from unrealized appreciation of equity securities such as SpaceX.

This "AI paper wealth" effect means that the index-level profit growth is more concentrated and more dependent on the revaluation benefits of a few leading companies than the surface data shows. Wall Street's pricing logic is therefore undergoing a subtle shift: the market is beginning to reduce rewards for simply "EPS exceeding expectations" and instead examines cloud business growth, order backlog, operating profit margin, and free cash flow after capital expenditure—paper revaluation can create a profit surprise, only customer payment and operating cash flow can support long-term valuation.

Rotation logic: from "shovel sellers" to "gold diggers"

In Tian Lihui's words, the market is shifting from "speculating on shovels" to "verifying gold mines." In the past two years, capital frantically bought stocks of "shovel sellers" like NVIDIA and Micron, firmly believing that selling shovels is the most profitable in a gold rush; but now, capital is flowing to "gold diggers" who really make money from AI—those who control data entry, paying users, and pricing power.

Behind this is the inevitable evolution of the AI industry cycle. When hardware expansion outpaces application demand,阶段性 overcapacity will emerge, and pure hardware manufacturing, if it loses demand support, can easily fall into price wars; the logic of cloud providers is relatively more solid because they have data, customers, and scenarios, and can convert computing power into sustainable cash flow. Many market judges believe this is not a few days of short-term fluctuation, but a landmark signal that AI is entering the application implementation period.

This rotation has also resonated at the macro level. US Treasury Secretary Bessent publicly stated that the US and Iran are likely to reach an agreement on opening the Strait of Hormuz in the short term, with international oil prices plummeting over 5% in a single day, with Brent crude falling below $80, and inflationary pressure and interest rate concerns cooling simultaneously. On the economic data front, the US July ISM Manufacturing Index rose to 55.6, the fastest expansion speed in more than four years, with AI data center construction becoming an important engine driving manufacturing and capital expenditure, providing fundamental support for the tech stock trend.

Implications for A-share and Southeast Asian investors

The sector rotation across the Pacific is transmitting emotions to global markets including A-shares. The Wind Semiconductor Index has adjusted from its high point of 11304.51 points on July 1 to around 7000 points, with the adjustment of domestic semiconductor stocks resonating with US stocks. Tian Lihui believes that in the short term, if global semiconductor demand cools, A-share and Hong Kong chip industries, especially storage and foundry links, will be suppressed in sentiment, and the risk of valuation killing needs to be警惕; but in the long term, external hardware cycle fluctuations will instead accelerate domestic substitution, while benefiting Chinese AI application layer companies with massive data and implementation scenarios.

For Southeast Asian investors, this differentiation also provides an important observation coordinate. First, the AI industry chain is not a monolith, and the simple logic of "buying tech stocks equals buying chips" needs to be corrected, as the profit realization capabilities of cloud services, AI applications, and enterprise software are differentiating; second, profit quality will become the core variable in the next stage of pricing, with the importance of indicators such as free cash flow and order backlog significantly increasing; third, the better-than-expected performance and sector rotation of the US earnings season will also indirectly affect the sentiment and valuation anchoring of Southeast Asian market tech sectors through capital flows.

Back to the investment strategy level, the consistent suggestion given by institutions is: don't pay for pure capital expenditure narratives, but use a magnifying glass to examine companies' free cash flow and main business growth rate. In the second half of the AI wave, it belongs to those companies that can convert computing power into profits and technology into customer payments—whether they are in Silicon Valley, Singapore, or Bangkok.

(This article is based on public market information and analysis, and does not constitute any investment advice. The market has risks, and investment requires caution.)