The AI investment cycle has entered a phase where returns are being put to the test. What will the next round of competition hinge on? Demand for AI continues to surge, but the price that tech giants are paying for this growth is becoming increasingly steep.
So far, five of the "Magnificent Seven" US stocks have released their latest quarterly earnings, and three of them reported negative free cash flow for the quarter: Amazon (NASDAQ: AMZN), Alphabet (NASDAQ: GOOGL), and Meta (NASDAQ: META) recorded net outflows of approximately $33.5 billion, $9.1 billion, and $2.1 billion respectively. Microsoft (NASDAQ: MSFT) and Tesla (NASDAQ: TSLA) achieved positive free cash flow of roughly $18.6 billion and $1.5 billion respectively. Apple (NASDAQ: AAPL) and NVIDIA (NASDAQ: NVDA) are expected to report their latest quarterly results next month.
Although capital expenditure and free cash flow accounting standards vary across companies, this divergence reflects that AI infrastructure investment is placing significant pressure on the cash flows of some tech giants. Following the earnings releases, Microsoft received positive market recognition for its better-than-expected cloud business growth and cash flow performance, while Alphabet fell 12.6%, marking its largest single-day drop since going public. Amazon's stock fell nearly 7% in after-hours trading following its earnings release, and Meta and Tesla shares also declined after their reports.
"In the past, the market was more willing to pay for AI's growth prospects and technology narrative, but the latest round of earnings from tech giants signals that AI investment is moving from an 'arms race' of capital expenditure scale into a phase of verifying commercial returns," Zhang Yi, CEO of iiMedia Consulting, told Yicai. Free cash flow has become an important indicator for measuring investment safety, but determining whether negative cash flow is dangerous still hinges on real orders, revenue realization capability, and sustained financing capacity.
Tech Giants Show Divergent Performance
The latest earnings from the three major cloud computing giants show similar growth trends but deliver different cash flow report cards: Amazon AWS, Microsoft Azure, and Google Cloud revenues grew 39%, 48%, and 103% year-over-year respectively. AI demand continues to drive cloud business expansion, but differences in capital expenditure, profit performance, and investment payback cycles are affecting how the market evaluates each company.
Amazon's latest earnings show third-quarter AWS revenue of $46 billion, up approximately 39% year-over-year, marking its fastest growth rate in nearly five years, with operating profit reaching $17.8 billion. The company disclosed that Amazon Bedrock spending by Fortune 500 customers grew nearly sevenfold year-over-year, and the number of tokens processed by the platform increased nearly 17-fold. The flip side of AWS's rapid growth is massive infrastructure investment. Over the trailing twelve months ending September, Amazon's free cash flow swung from an inflow of approximately $14.8 billion in the prior-year period to an outflow of $42.357 billion. The company expects cash capital expenditure of approximately $220 billion in 2026, with a further significant increase in 2027. The after-hours stock decline indicates that despite cloud growth exceeding expectations, it was not enough to fully offset investor concerns about cash burn.
Microsoft's performance was different. In its fiscal first quarter ending September, Azure revenue reached approximately $33.165 billion, up 48% year-over-year, higher than the 43% growth in the previous quarter. Microsoft's capital expenditure and finance lease-related investments for the quarter totaled approximately $45.2 billion, below the previously planned $50 billion; free cash flow reached approximately $18.6 billion. However, Microsoft's capital expenditure coming in below plan was not a deliberate pullback in AI investment, but rather delays in delivery of certain third-party computing facilities that affected the quarter's investment pace. The company's commercial remaining performance obligations reached $677 billion, up 73% year-over-year, providing some contractual support for future revenue; Microsoft expects Azure revenue growth to further accelerate to 49% to 50% in the next fiscal quarter.
Alphabet presented a more pronounced contrast between business performance and cash flow. Third-quarter Google Cloud revenue was approximately $30.7 billion, up 103% year-over-year; operating profit was approximately $11.1 billion, with an operating margin of 36.2%. Part of the growth came from external sales of Google's proprietary TPU systems, but the hardware business also requires advance procurement of equipment and inventory buildup, increasing working capital needs. Alphabet's capital expenditure for the quarter reached $51.4 billion, with free cash flow at negative $9.1 billion and full-year capital expenditure guidance of $195 billion to $205 billion. The company's cash flow pressure stems not only from AI infrastructure construction but also from increased TPU inventory and a $5.2 billion EU fine payment. Meanwhile, Google advertising revenue fell short of market expectations; net income declined approximately 25% year-over-year, mainly due to fair value losses on equity securities investments. These factors made Alphabet one of the tech giants with the most dramatic market reaction this earnings season. On October 29, Alphabet's stock fell more than 12%. For investors, cloud business growth rate is no longer the sole evaluation criterion; how quickly massive investments can be converted into profits and cash flow is becoming equally important.
Meta's business model differs from the three major cloud providers, but it faces the same balance between investment and returns. The company reported third-quarter revenue of $64.6 billion, up 26% year-over-year; operating profit was approximately $12.988 billion, down 37% year-over-year, with operating margin dropping from 40% to 20%. The decline was mainly affected by $9.89 billion in litigation-related expenses and cannot be entirely attributed to AI investment. Excluding that item, Meta's operating profit for the quarter was approximately $22.88 billion, up about 11.4% year-over-year. Meta's operating cash flow for the quarter was approximately $32.8 billion, capital expenditure was approximately $34.9 billion, and free cash flow was negative $2.12 billion, compared to positive $10.6 billion in the prior-year period. The company expects full-year capital expenditure of $135 billion to $145 billion in 2026. Meta's AI returns are mainly reflected in improved advertising efficiency rather than directly selling computing power. Incremental revenue from algorithms has already entered the advertising business, but future-oriented model, agent, and superintelligence research still requires sustained investment.
Tesla's third-quarter revenue was approximately $28.05 billion, roughly flat compared to the prior-year period; net income attributable to common stockholders was approximately $790 million, down 43% year-over-year, with free cash flow of approximately $1.53 billion. The company also revised its 2026 capital expenditure expectation from over $25 billion to over $20 billion. However, Tesla's capital expenditure for the first three quarters was still significantly higher than the prior-year period, and the company continues to invest in AI computing power, robotics, and manufacturing facilities. Additionally, the potential merger between Tesla and SpaceX that is being evaluated further adds complexity to its capital allocation.
What Will the Next Round of Competition Be About?
The latest round of earnings at least shows that global tech giants have not noticeably slowed AI infrastructure construction. Amazon expects capital expenditure of approximately $220 billion in 2026, Alphabet expects $195 billion to $205 billion, Meta expects $135 billion to $145 billion, and Microsoft slightly reduced its guidance to $170 billion. The capital expenditure scale of these companies has reached unprecedented levels, and industry estimates suggest next year's total could reach $1.3 trillion. Research published by S&P Global Ratings this year estimates that six hyperscalers it covers (Alphabet, Amazon, Microsoft, Meta, Oracle, and SpaceX) will post negative free operating cash flow in 2026 and 2027, with cash flow expected to recover only by 2029. The agency also listed AI investment monetization capability, demand sustainability, overcapacity risk, and debt-like contractual commitments as key monitoring areas.
However, negative free cash flow does not directly mean investment failure. Data centers, AI servers, and networking equipment typically require concentrated cash payments during construction and procurement phases, with costs gradually recovered through cloud services, model inference, and enterprise applications over subsequent years. An AI industry practitioner told reporters that negative free cash flow for a single quarter or even several consecutive quarters may reflect an investment cycle rather than necessarily indicating an unsustainable business model. What truly needs to be observed is whether future revenue growth can cover equipment depreciation, operating costs, financing expenses, and the next round of hardware upgrades.
"The key to the next round of competition is not how much computing power is invested, but whether computing power can be converted into profits on financial statements," Zhang Yi said. Unlike previous internet and mobile internet development stages, in the AI heavy-asset era, two indicators deserve particular attention: first, order visibility, such as whether remaining performance obligations (RPO) and backlog orders can provide future revenue support; second, capital discipline, meaning whether companies can maintain necessary investment while safeguarding cash flow底线, ultimately forming a positive cycle from computing power investment to product and application deployment to commercial monetization.
Even with orders in hand, tech giants still need to solve funding and asset management challenges. Companies with stable cash flow, higher credit ratings, and long-term customer contracts can obtain funding at relatively low cost and lock in data centers, power, and chip resources in advance; but the more complex the financing structure, the more investors need to carefully assess a company's true leverage and long-term commitments.
Technological iteration can also change investment returns. AI chips and servers update rapidly, and new-generation products often significantly outperform previous-generation equipment in performance and energy efficiency. For companies investing tens of billions of dollars in building data centers, computing power utilization rates, actual economic life of equipment, and depreciation schedules will all affect future profits; if hardware updates faster than expected, companies may also face higher depreciation or asset impairment pressure.
The market is not uniformly pessimistic about AI investment returns. The rapid growth of AWS, Azure, and Google Cloud already shows that enterprise computing power demand is still expanding. As model inference and enterprise AI application usage increase, existing infrastructure may improve returns through higher utilization rates. Microsoft sells AI services through Azure and enterprise software, Amazon relies on AWS and Bedrock to provide computing power and model services, Alphabet simultaneously advances cloud services and TPU commercialization, and Meta mainly uses AI to improve advertising efficiency. Investment payback cycles differ across business models, but these businesses have already provided real demand support for massive infrastructure investment.
Next, NVIDIA will release its new quarterly earnings in November, providing fresh evidence for observing upstream computing power demand. The AI investment race continues. As more data centers and computing facilities come online, the market will gradually see just how much of this unprecedented capital expenditure can be absorbed by continuously growing AI demand.
Comments