Tech giants are engaged in an unprecedented capital expansion, but market patience for this high-stakes bet is diverging. This earnings season has revealed a core contradiction: the scale and speed of AI investment far outpace the pace of revenue realization. While strong cloud business growth has bought some companies breathing room, the collective deterioration of free cash flow has weighed on the stock prices of others. The outcome of this narrative battle will become clear over the next several years.
In the latest quarter, Amazon (AMZN.US) capital expenditures surged 69% year-over-year to approximately $54 billion, Meta (META.US) capital expenditures rose 83% to $31.1 billion, Microsoft capital expenditures also climbed 69%, and Google raised its full-year spending target by $15 billion to a maximum of $205 billion. According to Wall Street estimates compiled by FactSet, the combined capital expenditures of these four companies are projected to reach roughly $1.6 trillion over the next two years. Since the AI race began in 2023 through June of this year, their cumulative capital investment has already exceeded $1.1 trillion.
Market reactions have been sharply divided. Amazon's cloud business posted its fastest growth rate since early 2022, sending its shares up nearly 10% in after-hours trading. Microsoft Azure achieved a record single-quarter sales figure, with its stock rising over 15% the next day. In contrast, Meta saw its shares fall more than 7% due to a disappointing revenue outlook and continued spending expansion. Following its first negative free cash flow quarter since going public, Google shares dropped over 6% the next day. The combined free cash flow of the four companies fell to a decade-low of approximately $7 billion.
Spending Scale: An Arms Race Without a Finish Line
Looking at quarterly data, the combined capital expenditures of the four companies reached $170 billion in the second quarter, a year-over-year increase of about 78%. This figure is being revised upward each quarter. Amazon raised its full-year capital expenditure forecast by $20 billion to about $220 billion, primarily due to a sharp rise in memory chip prices. Google increased its annual target by $15 billion, and Meta raised the lower end of its full-year capital expenditure range from $125 billion to $130 billion. The continuous expansion in spending is partly driven by supply chain pressures. Reports indicate that a shortage of memory chips has pushed up procurement costs. Microsoft disclosed in April that soaring chip prices would add an extra $25 billion to its expenses this year. This cost pressure has even affected Apple, which is not directly involved in the AI race. The company warned that rising costs would suppress sales and profit margins, causing its stock to fall 6.3%. RBC Capital analyst Rishi Jaluria noted, "The growth in capital expenditures essentially has no end in sight. Investors need these companies to walk a tightrope between AI investment and maintaining the competitiveness of their existing businesses."
Cloud Computing: The Core Channel for AI Monetization
Currently, the strongest support for the "investment is effective" narrative comes from the accelerating growth of the cloud computing business. Amazon's cloud unit, AWS, saw second-quarter sales rise 37% to $42.2 billion, with operating profit also improving. Microsoft Azure's full-year sales surpassed $100 billion for the first time, with the single-quarter growth rate being the fastest since 2022. Google Cloud added approximately $11 billion in revenue year-over-year. The cloud customer base for these three companies spans from OpenAI and Anthropic to various enterprise users. The contract backlog driven by AI computing demand has become a key indicator supporting growth expectations. The combined contracted backlog for Amazon, Google, and Microsoft is nearly $1.7 trillion, more than doubling from a year ago. Amazon CEO Andy Jassy stated on the conference call that cloud services "could well become our trillion-dollar-a-year revenue business in the future." Microsoft CEO Satya Nadella noted that the company opened 31 data centers across five continents last quarter and will double its overall computing capacity within two years.
AI Revenue: From Spending Narrative to Monetization Validation
In this earnings season, the four companies responded more quantitatively than ever to the market's core question: "The money is spent, but how much has been earned back?" Amazon was the first to provide the most specific figures. Its AI business within AWS has an annualized revenue run rate exceeding $25 billion, and its self-developed chips (Trainium and Graviton) also surpassed $25 billion, with both achieving triple-digit percentage year-over-year growth. Jassy highlighted this as a key disclosure highlight of the earnings report, signaling that Amazon is beginning to separate AI revenue from overall cloud business data for market scrutiny. Microsoft's AI annualized revenue run rate reached $37 billion. Combined with Microsoft 365 Copilot paid seats surpassing 30 million (up from about 20 million three months ago, a roughly 50% quarter-over-quarter increase), it has formed a dual monetization structure of "cloud AI + office AI." Nadella emphasized, "We are pushing the cost-efficiency curve to new frontiers, ensuring every customer can convert tokens into tangible business outcomes." While Google did not break out AI revenue separately, management confirmed that Vertex AI, the Gemini API, and TPU compute leasing were the core drivers of its cloud business's 82% growth rate. The Gemini API now processes 22 billion tokens per minute, up from 16 billion last quarter, and Gemini Enterprise is used by nearly 90% of Fortune 100 companies. Google Cloud's backlog surged from approximately $460 billion to $514 billion, an increase of over $50 billion, primarily driven by enterprise AI demand. The company expects more than 50% of this to convert into revenue within the next 24 months. Meta's AI monetization path is entirely embedded within its advertising engine. The AI-driven Advantage+ advertising tool has an annualized revenue run rate of about $75 billion, while AI value optimization tools have an annualized run rate of over $20 billion (doubling within a year). AI improvements to recommendation systems and ad conversion rates led to a 12% year-over-year increase in ad unit prices and a 14% increase in ad impressions, driving a 27% year-over-year increase in family of apps ad revenue to $59.4 billion. CEO Mark Zuckerberg stated on the call, "AI is accelerating our core business today." However, lacking a standalone cloud business as a direct monetization channel, Meta's AI investment return is more difficult for the market to price compared to the other three.
Free Cash Flow: Consensus on Pressure, Disagreement on Guidance
Despite revenue growth, the deterioration of free cash flow is a common theme in this earnings season that cannot be ignored. Amazon's operating cash flow over the past 12 months was $161.4 billion, up 33% year-over-year, but after deducting investments in data centers and other infrastructure, free cash flow plummeted to negative $7.6 billion. Google recorded its first negative free cash flow quarter since going public over 20 years ago, with a deficit of about $6 billion. The combined free cash flow of the four companies fell to approximately $7 billion, the lowest in a decade, with only Microsoft and Meta achieving positive figures. Jassy acknowledged that while simultaneously building multiple data centers, with a lag of about two years from project initiation to launch, "in the short term, we will spend a lot of capital and face free cash flow pressure until these data centers come online." Dec Mullarkey, Managing Director at SLC Management, said, "For investors, the era of growth at any cost is over. They want to see spending translate into performance, like Google, Microsoft, and Amazon." This partially explains why Meta's stock faced greater pressure after its earnings report.
Concentration Risk: The Butterfly Effect of OpenAI and Anthropic
Another hidden risk in the current capital expenditure frenzy is that a portion of demand is highly concentrated among a few AI startups. The significant growth in the contract backlog is largely due to multi-year compute purchase commitments from OpenAI and Anthropic. This means that the tech giants' investment returns partly depend on these two companies' ability to continue raising funds, fulfill contracts, and execute their respective IPO plans. Moody's has warned that this structure creates "a more cyclical system that could mask true demand," posing a concentration risk that cannot be ignored. Meanwhile, disclosures of future financial commitments have revealed deeper balance sheet lock-in risks. According to the Financial Times, in the second quarter alone, Meta, Google, and Microsoft signed nearly $900 billion in new AI-related obligations, covering data center leases, infrastructure procurement commitments, and new debt. Meta also added an additional $68 billion in data center leases in July. Amazon has not yet disclosed similar detailed data.
Narrative Divergence: Who Wins, Who Waits
After this earnings season, the market's criteria for evaluating tech giants' AI investments are becoming more specific. It's not just about "how much they spent," but more importantly, "how much they earned." Amazon and Microsoft have temporarily earned market tolerance for high capital spending, thanks to the accelerating growth of their cloud businesses. Google has raised investor concerns due to its cash flow issues, despite its cloud business also performing well. Meta's situation is more complex. While AI-powered ad targeting drove total revenue up 28% year-over-year to $61 billion, the lack of a direct cloud-based monetization channel, combined with CEO Mark Zuckerberg's comments about renting out data center compute capacity, which were seen as lacking a clear plan, led to a negative market reaction. Alphabet CFO Anat Ashkenazi stated on an analyst call last week, "As long as we see these attractive investment opportunities, we will continue to invest." This statement reflects the common position of the management teams of all four companies. However, as RBC's Jaluria noted, "Investors are being forced to reassess their own timelines." The race between capital expenditure and AI revenue is still far from being decided.
This article was originally published by "Wall Street CN" and edited by Li Fo for Zhitong Finance.
Comments