The highlights of the financial reports of the four major technology giants, Google, Microsoft, Amazon, and Meta, will be released after the US stock market closes next Wednesday. This should be the most important day of this financial reporting season, far more important than the financial reports of other companies or even Nvidia. There is nothing else, the core is that the financial reports of these four companies, especially the revenue from cloud business (meta is advertising), are a key indicator for the market to verify whether AI commercialization revenue has reached a turning point. The market is speculating on the computing power industry chain, and ultimately downstream revenue is needed to support it. This quarter needs to show the market that such a large capital expenditure can be a profitable and super profitable start. Otherwise, doubts about capital expenditures will return. Several highlights of the core financial reports of the four major tech giants from a personal perspective 1. From the perspective of performance growth rate: Azure 38%, AWS 32%, GCP 58%, Meta Ads 32% - Backlog release rhythm determines slope. UBS provided the most complete caliber on the 23rd, and the growth guidance for the three public clouds in 1Q26 is as follows: 1) Microsoft Azure: 38% (constant exchange rate), core driver 32% of enterprise GenAI expenses are accounted for. Morgan Stanley maintains a target price of $650 on 4/14, citing that there is the greatest potential for valuation repair during the AI transformation period. 2) Amazon AWS: 32% (America's radical version can reach 38% for the whole year, far exceeding the consensus of 26%). Anthropic contract revenue increased fivefold from a monthly $6 billion to $30 billion in the first half of the year, and non-linear acceleration has been confirmed. 3) Google Cloud (GCP): 58% (Citi 4/14 raised to 57.5%, corresponding to an OI profit margin of 34.6%). The denominator is the smallest (4Q25 quarter revenue of approximately $15 billion vs AWS $29 billion), but the proportion of AI long contracts exceeds 70%. The estimated shipment volume of TPU v6e/v7 in 2026 is 4.58 million pieces (an increase of 168% compared to 2025), and 1Q-2Q will be released in a concentrated manner. 4) Meta: Advertising business forecast of $56.1 billion, year-on-year 32.6%. Advantage+, WhatsApp advertising, and Reels ARR of 50 billion+have entered a "high-level platform" with stable growth, forming two AI monetization rhythms in contrast to GCP's "just entering acceleration curve". The source of the growth rate difference is not simply the "base effect", but three nested variables: 1) order structure (GCP AI long contracts account for a higher proportion, AWS traditional IaaS dilution slope); 2) Capacity unlocking rhythm (GCP TPU concentrated volume); 3) Absolute incremental (AWS still leads by $7.5 billion vs GCP by $5.5 billion). Verifiable judgment: If the "deferred revenue+RPO quarter on quarter growth rate" given by Google's financial report is lower than 8%, the sustainability of the overall GCP58% growth rate may be falsified Meta Reality Labs' 1Q suffered a loss of 4.7 billion yuan, with a peak of 19.2 billion yuan for the whole year. The fluctuation will be amplified, but it will not change the nature of the advertising AI high-level platform. 2. Capital expenditure perspective: Who is spending $660 billion, where is it being spent, and when does depreciation begin to drag on Summary of Capex Guidelines for Four Companies in 2026 (Figure 2 below) 1) Microsoft: FY27 Citi predicts $192 billion (14% higher than consensus), with buyer research even assuming a 40% growth. 2) Google: $175-185 billion (Citi 4/14), $226 billion (+25%) in 2027, announced infrastructure exceeding $140 billion (including $40 billion in Texas, $15 billion in Indian submarine cables, Wiz acquisition). 3) Amazon: Approximately $200 billion, primarily investing in AWS AI infrastructure. 4) META: $115-135 billion, while external long-term lease contracts (Google Cloud $10 billion, CoreWeave $19 billion+, Oracle $20 billion, AMD $100 billion/6GW, etc.) totaling over $176 billion, with depreciation transferred. Divided by AI data center structure: GPU/ASIC accounts for 50% (325 billion), building+power accounts for 25% (162.5 billion), network accounts for 15% (97.5 billion), and storage accounts for 10% (65 billion). The non GPU portion is disproportionately amplified - Nvidia's data center revenue corresponds to a decrease in the proportion of general GPU purchases from 35% in 2024 to 22-25% in 2026, spreading to Broadcom AVGO (custom ASIC), SK Hynix (HBM), Arm, Bloom Energy, and others. Depreciation turning point: UBS predicts that there will be a significant delay starting from the second half of 2027 (with an annual increase of approximately 100 billion yuan in depreciation, based on a 5-6 year cycle). The most noteworthy hidden indicator is the "revenue/Apex ratio", which measures the core indicator of depreciation inflection point and is estimated to be around 4.0 in FY25; If it falls below 3.5, the market pricing will not be able to outperform depreciation due to capacity ramp up; Hold on to 3.8+and postpone the turning point to at least the first half of 2028. 3. Angle of remaining unfulfilled orders: $1.7 trillion backlog (orders signed but not fulfilled), UBS provides a panoramic view of the order book of large cloud vendors by the end of 2025 (Figure 3): MSFT 625 billion (AI accounts for 30-35%); AWS+GCP AI accounts for over 60%; ORCL surged to 523 billion mainly due to a single large order from OpenAI. After divesting from traditional contracts, the pure AI computing power order book is about 800-900 billion US dollars, corresponding to an annual equipment demand of 60-70 billion US dollars for 5-7 years, which is commensurate with Nvidia's data center revenue volume in 2025. A $1 order corresponds to approximately $0.4-0.5 GPU/ASIC procurement, with the remaining being infrastructure. Pay attention to the three dimensions of the value of the backlog: 1) whether the cash flow of the demand side and the contract side is stable (OpenAI, Anthropic, Apple, Meta vs 2022 SaaS startups); 2) Mandatory structure (booking+advance payment, high breach of contract costs vs. "usage commitment"); 3) Real consumption verification (Anthropic monthly ARR is 5 times higher than half a year). Highlights of Q1 2021 Financial Report Update: RPO is likely to increase by 20-25%. Focus on the "weighted average remaining term" - extend it to 5.0 years or more to confirm the "long contract era"; If shortened to less than 3.8 years, the short-term contract will flow back and the pricing needs to be reassessed. In summary, this round of consensus revision+Apex/Backlog synchronization explosion is a key turning point in the AI infrastructure cycle from "expectation driven" to "fulfillment driven". As long as two of the three indicators of RPO month on month, revenue/Apex ratio, and Cloud guidance do not collapse, the sector will still maintain a positive feedback of "upward revision cash out further upward revision". But high expectations are already in place, and any 'flat guidance' could trigger a short-term pullback. This means that the risk on the day of the financial report is that the guidance is not aggressive enough - buyers have pushed their expectations higher than sellers (JPMorgan Chase CY27 capex is 14% higher than consensus). If any company's FY27 guidance only reaches a "flat consensus", the market will react with a "lower than implied caliber", repeating the script of multiple previous financial reports indicating a weak market followed by a decline. This article is sponsored by @ bitget_zh, titled 'Bitget Buying US Stocks: Instant Entry, Smooth Trading'