How to interpret Google's Q3 earnings report? One highlight and two concerns. A cursory glance at the lower-than-expected revenue and profit figures might seem alarming, but a closer look reveals that the cloud business, GCP, significantly exceeded expectations. The disappointing aspects were the underperformance of search advertising and YouTube advertising, which dragged down the overall group's revenue and profit. Let's discuss in detail: 1. Highlights of the Cloud Business: In Q3, Google Cloud revenue reached $30.709 billion, a year-over-year increase of 102.6%; operating profit was $11.121 billion, with an operating margin of 36.2%. Building on nearly $25 billion in revenue in Q2, this represents a sequential increase of approximately 24%, demonstrating strong expansion. Of course, this includes TPU sales. Management (including CFO Anat Ashkenazi and Google Cloud CEO Thomas Kurian) explicitly stated that GCP contributed the vast majority of Q3 cloud revenue; excluding the impact of TPU system sales, cloud revenue growth was "again incredibly strong." Therefore, it can be said that even excluding TPU sales, GCP's growth rate is very strong. This is likely also supported by the accelerating commercialization of AI. 2. One concern is the core business underperforming expectations. While cloud business growth is strong, search advertising and YouTube advertising still account for the majority of the group's revenue. Underperformance in these areas raises concerns that the core business is bleeding money. The slowdown in search advertising has been explained by four factors: exchange rates, the impact of new EU regulations on some retail advertising, the fading of the World Cup advertising boom in June, and the monetization of AI Mode. The low monetization of AI Mode is the most noteworthy structural factor, which has been acknowledged by management. However, the official response has not yet addressed several crucial questions for long-term value: 1) Does AI Mode primarily increase queries that wouldn't have occurred previously, or does it replace existing queries? 2) How many of the migrated queries originally carried strong purchase intent? 3) After expanding advertising coverage, to what level can revenue per query recover? 4) After deducting model inference costs, how much profit can recover? The market has already seen the impact of AI on Google Search advertising monetization, with the growth in AI search usage outpacing advertising commercialization. However, the lack of clear and explicit explanations for the above four points naturally increases concerns. 3. Another concern is capital expenditure and cash flow. Third-quarter free cash flow fell to negative $9.07 billion, far exceeding market expectations of negative $3.5 billion; capital expenditure was $51.36 billion, lower than market expectations of $55.78 billion, maintaining the guidance of $195-205 billion in capital expenditure for the full year 2026. However, free cash flow needs closer examination due to two unconventional factors: 1) TPU system inventory buildup: Inventory increased from approximately $2.4 billion at the end of 2025 to approximately $13.3 billion at the end of the quarter. Q3 inventory cash outflow was approximately $3.2 billion (cumulative over nine months approximately $11 billion). This is for building up inventory for TPU systems that have been contracted but not yet delivered/revenue recognized; cash outflow precedes revenue recognition, representing a "different cash flow cycle." 2) European Commission fine cash payment: Approximately $5.2 billion (paid after the appeal of the relevant fine in 2018), directly reducing operating cash flow. If inventory outflow and fines are added back, the adjusted figures will improve significantly (estimated to be close to flat or slightly negative). The significant drop in profit and EPS compared to expectations necessitates consideration of the impact of unrealized losses on equity investments, primarily unrealized (book) losses on SpaceX. 1) Net losses on equity securities were approximately $8.66-8.7 billion, reducing net profit by approximately $6.7 billion and EPS by approximately $0.55. 2) Other income and expenses (OI&E) were -$8.2 billion. The CFO explicitly stated that this was "primarily due to unrealized loss from our investment in SpaceX" (mainly due to unrealized losses from the quarterly decline in SpaceX stock price). This is a paper loss measured at market capitalization, not a cash loss from actual stock sales. In summary: Profits were dragged down by unrealized losses from SpaceX (non-cash); the negative FCF depth was mainly due to high CapEx + cash outflows from TPU inventory preparation + cash payments for EU fines; CapEx was lower than expected in the quarter, but full-year guidance remains unchanged; the company is continuing to finance AI capital expenditures through ATM equity issuance and bond issuance. Returning to the market impact: The market saw strong performance in the cloud business, but weaker-than-expected search performance deepened concerns. Worse-than-expected free cash flow raised concerns about the sustainability of Capex (although two of the reasons are unconventional). However, the slowdown in the core business is a real phenomenon, and Google continues to raise ATM funding (with nearly $30 billion in unused funds), leading to a sharp drop in after-hours trading. My personal view is that Google's core competitiveness remains: rapid growth in its cloud business (this is key), a full-stack AI strategy, and over $200 billion in cash reserves. The market needs to digest the negative impact of the core business slowdown and the fact that the Gemini flagship model hasn't fully caught up. A reversal will require seeing AI-mode search commercialization catch up with usage, or seeing the flagship model and agents catch up.