In July, the U.S. stock market witnessed a wave of massive collective bets on AI, with tech giants planning to invest a total of nearly 800 billion U.S. dollars throughout the year.

Tech giants are ramping up investments in AI infrastructure at a frantic pace, and "cash burning" has become the only core theme in the current US stock technology track.

Throughout July, the seven leading US stock tech giants, AI computing power and semiconductor enterprises presented a highly unified yet sharply differentiated market trend. At the beginning of the month, the sector collectively pulled back, the Philadelphia Semiconductor Index plummeted 11% in a single week, nearly 70% of semiconductor stocks retreated by more than 20% from their recent highs, as investors worried that massive capital expenditures would continuously erode corporate cash flow.

After the concentrated earnings disclosure at the end of the month, the market trend reversed completely. Microsoft surged 15.51% in a single day, marking its largest single-day gain since 2008, directly driving a full-line counterattack in storage, equipment, and AI chips. The Philadelphia Semiconductor Index rose 8.19% in a single day, with Micron, Western Digital, and AMD all posting single-day gains of over 13%.

01

All leading enterprises share a high degree of commonality in their core business actions in July:

They keep raising their full-year capital expenditure, with almost all funds flowing to AI servers, data centers, self-developed chips and storage production capacity. According to rough statistics from Jiemian News, large technology companies including Amazon, Alphabet, Microsoft, Meta and others have disclosed concentrated full-year capital expenditure plans of nearly 800 billion US dollars in a single month.

Alphabet, Google's parent company, took the lead in raising its full-year capital expenditure to 195-205 billion US dollars in mid-July. Google CFO Anat Ashkenazi said on the earnings call that Google Cloud's backlog of orders exceeds 514 billion US dollars, and the computing power gap will last for at least two years.

Microsoft maintained its original capital expenditure plan of 190 billion US dollars, but its quarterly capital expenditure reached 41 billion US dollars, surging 70% year on year, all of which was used to purchase GPUs to expand the Azure computing power cluster. The company's paid Copilot users continue to grow at a high speed, and the investment has simultaneously generated revenue, which is the core support for Microsoft's skyrocketing stock price at the end of the month.

Meta and Amazon released signals of increased investment intensively at the end of July. After releasing its Q2 earnings report, Meta raised its full-year capital expenditure guidance from 125 billion US dollars to 130-145 billion US dollars, with quarterly capital expenditure reaching 31.08 billion US dollars, a sharp increase year on year.

Meta CFO Susan Li said on the conference call that the industry will face a long-term shortage of computing power supply, and the company will maximize the speed of data center construction from 2026 to 2027, and simultaneously promote the 14-billion-US-dollar supercomputing park project in Texas and the 500-billion-US-dollar supercomputing park project in Louisiana.

Unlike other cloud giants, Meta does not have a cloud business that sells computing power externally, and all its AI investments can only be digested through monetization of advertising. Its free cash flow plummeted 91% year on year to 784 million US dollars in Q2, and its stock price fell 8% in intraday trading after the earnings report was released, making it a weak stock in the July capital expenditure expansion market.

Amazon also released a major expansion plan. Amazon CEO Andy Jassy announced at the earnings meeting on July 30 local time that the full-year capital expenditure expectation has been raised again to 220 billion US dollars, 20 billion US dollars higher than the initial expectation in February.

Jassy judged that the imbalance between supply and demand of computing power will continue until 2028. "Even at such a scale of expenditure, we still cannot meet all the computing power demand in 2026, and the demand in 2027 and 2028 will remain strong."

Amazon AWS's cloud business revenue in Q2 reached 42.2 billion US dollars, a 37% year-on-year surge, hitting the highest level in 18 quarters. The annual revenue of its AI business exceeded 25 billion US dollars, achieving triple-digit growth. Sufficient cash flow hedges the pressure of capital expenditure. After the earnings report was released, Amazon's after-hours stock price surged by nearly 10%, forming a clear divergence with Meta.

Hardware leader Apple also released its latest earnings report on July 30 local time. In the third fiscal quarter of fiscal year 2026, the company's hardware revenue increased across the board, but the production capacity guidance for the fourth quarter was weak, coupled with the service business growth rate falling short of expectations, the stock price plummeted 6.33% after hours following the earnings release.

Although Apple has not built large-scale self-operated AI data centers, it continues to increase investment in the R&D of self-developed on-device AI chips, and simultaneously locks in long-term supply of storage and advanced chips from upstream suppliers. This is essentially an extension of the logic of upstream cost expansion in the AI industry chain, which echoes the computing power expansion of other giants at the upstream and downstream of the industry.

In addition, Tesla reiterated in July that its full-year capital expenditure will exceed 25 billion US dollars, and its self-built AI chip factory in Texas has been put into operation, which mainly relies on autonomous driving on-device AI to achieve computing power expansion.

02

As direct beneficiaries of capital expenditure, semiconductor enterprises fully aligned their business operations in July with those of tech giants, forming a closed loop of supply and demand.

Intel released its earnings report on July 24, with quarterly revenue of 16.1 billion US dollars, up 25% year on year, and revenue from data center AI business surging 59% year on year. Its full-year capital expenditure was raised to 20 billion US dollars. The company is ramping up investment in advanced process factories in the United States, and has signed ten long-term server CPU supply agreements with cloud vendors to lock in orders.

NVIDIA's stock price fluctuated and weakened in July, falling to a low of 190 US dollars during intraday trading. The market worried that the continuous expansion of cloud vendors would lead to excess chip supply, but the end-market demand data remained strong. It rebounded along with the broader market at the end of the month, closing up 2.65%, with a total market capitalization remaining at a high level of 4.7 trillion US dollars.

AMD, storage vendors Micron, Western Digital, and equipment enterprise Applied Materials all received large long-term orders from super-large-scale cloud vendors in July. Storage chip enterprises benefited from the incremental storage demand brought by AI computing power, and ushered in a collective surge at the end of the month, with SanDisk posting a single-day gain of nearly 26%.

For this round of computing power expansion, some analysts pointed out that capital is rotating from the seven giants to the semiconductor track. The market rewards the beneficiaries of capital expenditure while questioning cloud vendors that keep burning money, and the short-term differentiated market trend will continue.

BlackRock, the world's largest asset management company, believes that this round of sell-off is "overreaction", saying that the market has confused "the transformation of the AI competitive landscape" with "the collapse of AI investment". BlackRock emphasized that cheaper AI will not end investment, but will instead accelerate the popularization of applications across various industries, expand the potential market size, and then drive up the overall demand for data centers, storage chips, power and other infrastructure. The institution maintains an overweight stance on US stocks, and recommends focusing on AI bottleneck fields such as chips and power.

This article is from "Jiemian News", reporter: Song Jianan, published with authorization from 36Kr.