#EarningsObserver: Amazon's guidance falls short of expectations, yet the stock price rises 9% Amazon $AMZN's net profit this quarter includes approximately $53.4 billion in pre-tax non-operating income Operating profit related to the core business is $27.5 billion, a 43% year-over-year increase This portion of profit is mainly driven by AWS, with Q2 revenue reaching about $42.2 billion, a 37% year-over-year growth, the fastest pace in the past 18 quarters The growth curve in Chart 2 forms an "inverted U" shape, with the latest data just surpassing Q1 2022 by 0.2 percentage points 🤣 AWS accounts for about 21% of Amazon's total revenue but contributes approximately 60% of operating profit Regarding this data, I believe the logic is the same as the Microsoft earnings viewpoint I wrote a few days ago: both have entered the acceptance phase empowered by AI Neither has stopped investing in expenditures; this fastest growth also coincides with Amazon's heaviest capital expenditure and most obvious free cash flow pressure Therefore, the significance of AWS's growth this time is not just the cloud business recovering growth, but it preliminarily validates Amazon's capital expenditure direction Previously, AI sector concept stocks or those related to AI were almost all in the "investment-first" stage, differing only in whether they were opening new businesses or adding on existing ones But regardless of the type, there is a core contradiction: the speed of demand realization One data point: AWS backlog increased from about $364 billion last quarter to $496 billion, a single-quarter increase of about $132 billion Although backlog cannot be directly equated to future revenue, it at least indicates that AWS currently faces clear market demand; the remaining issue is whether capacity can be continuously realized That's why I compared it with the Microsoft earnings viewpoint I posted a few days ago: capital expenditure without clear demand is a bet, but the trend of accelerating realization has begun, which is a story of another stage Also, from Chart 2, you can see that the growth trend in the previous two quarters was quite strong, but why is the reaction particularly large this time? I think one reason is market sentiment, causing the market recently to shift focus from AI investment scale to whether the investment can truly be realized The second is that the same number has completely different meanings under different operating backgrounds Although AWS's growth rate was also strong in the previous two quarters, the market mostly saw it as "cloud business recovery" This time, the 37% growth rate appears alongside several data points: ➠ Free cash flow turned from positive to negative ➠ Capital expenditure continues to rise ➠ Backlog increased by $132 billion in a single quarter ➠ AWS operating margin further rose to 39.4% Putting these data together, the market understands not just "cloud business is growing well," but that Amazon's revenue side is accelerating during its heaviest capital expenditure period, and the profit side is not obviously dragged down by new investments The first two quarters confirmed the trend; Q2 is more like a threshold breakthrough Looking at the market (Chart 3): The post-earnings gap up has already factored in some narrative repair I generally do not recommend building positions around earnings releases because trading at this stage involves not only fundamentals but also expectation gaps and sentiment From the daily K chart, Amazon's price has reached near previous high levels, and the RSI shows a clear short-term overbought condition So a simple conclusion can be given first: around $270 is not a position to immediately chase after a post-earnings gap up Fundamentals are worth watching, but fundamentals and price suitability for immediate entry are two completely different things

The Perp DEX quarterly report posted the day before yesterday Although the four selected protocols cannot represent the entire Perp DEX industry, they basically form an operational sample from the top, middle, to the tail end #交易之声：你的经验值得被听到 Their combined Q2 revenue decreased by 21.4% quarter-over-quarter, and this was under conditions where Q2 volatility was more pronounced than in Q1 I also checked the derivatives trading data of CEX, and after statistics, it was also found to have declined quarter-over-quarter Above is the aggregated revenue decline situation, shown in the data of Figure 2 Although all four protocols declined, the proportions vary greatly, with the middle two having the largest declines, both over 50% This means that when market trading activity decreases, users and liquidity do not withdraw evenly from all platforms It is inferred that funds may have first left platforms with weaker user stickiness or those reliant on incentives, and only later affected the leading platforms with stronger network effects ➠ Or put another way, the mid-tier protocols might be considered backups, and when the "market temperature" drops, the "spare tires" are more easily discarded The tail-end protocols, due to their relatively small user base, are not as heavily impacted as the mid-tier protocols with a certain scale _____________________________________ Then looking at profit margins, we can temporarily use 70% as an observation line, dividing the four Perp DEXs into two basic tiers: Hyperliquid @HyperliquidX and Lighter @Lighter_xyz have profit margins exceeding 70%, while edgeX and ApeX are below 70% This line reflects not only the platform's bargaining power but also cost control, liquidity efficiency, and ecosystem incentive expenditures, so it cannot alone represent complete operational quality To further assess a Perp DEX's operational quality, three dimensions need to be observed simultaneously: revenue scale, profit margin stability, and costs Cost items vary somewhat by protocol but basically revolve around these elements: market maker rebates, referral rewards, liquidity incentives, and other ecosystem allocations These expenditures are even sometimes not positively correlated with trading volume ➠ For example, liquidity: even if a trading platform's quarterly trading volume declines, it must maintain order book depth, market making quality, and user experience, otherwise revenue decline will further evolve into user loss The original report also included my views on these four Perp DEXs from the above three dimensions, but the main chapter is too lengthy to include here However, a few brief points are necessary: 1. Lighter's profit margin rose rather than fell, increasing from 74.6% in Q1 to 75.6% in Q2; it is also the only platform among the four protocols to improve profit margin despite revenue decline It is also the only protocol in Q2 to surpass Hyperliquid in profit margin 2. edgeX had the largest profit margin decline among the four protocols; although its costs also decreased, the problem is that cost contraction lagged significantly behind revenue decline This phenomenon can be understood as negative operating leverage _____________________________________ Finally I temporarily divided the Perp DEX sector into three tiers: ➠ The first tier consists of protocols that have already formed operational barriers, currently only Hyperliquid; ➠ The second tier includes protocols with decent operational efficiency and some operational quality but market share in the second echelon, such as Lighter and edgeX (the former's operational quality belongs to the first echelon); ➠ The third tier includes platforms with value capture mechanisms but insufficient business scale, namely ApeX, which has accumulated a large stock of Perp DEXs. I think this is a good phenomenon; although the first-tier Hyperliquid occupies a considerable market share, it does not resemble the oligopoly pattern of re-staking or oracle sectors Nor is it like DEXs with homogeneous and evenly divided shares The growth curve of mid-tier Perp DEXs shows high vitality Currently, such sector structures are rare in the market

The earnings reports released this time by $MSFT and $META are not "one good report and one bad report." On the contrary, both of their revenues have grown, but in the latest after-hours trading, one is going up while the other is going down. #财报观察员：微软云收入破千亿，Meta却指引拉胯——AI故事分化了？ In an earlier short post I published, I mentioned: I believe the issue lies in cash flow. Because if you look at revenue and capital expenditures, the results they present to the market are similar: both revenue and capital expenditures are growing. In this article, I will further break down the specific data and business structure to see if cash flow is indeed the "mirror" answer. Microsoft has been continuously investing in the AI field over the past few years, which can be considered a long-term strategic layout. From this earnings report, this investment is receiving feedback on different levels and forming a closed loop from the underlying infrastructure to the end user: At the base layer, Azure supports the growing demand for cloud computing and AI computing power from enterprises; At the application layer, the user base and usage of Copilot continue to expand; Ultimately, these growths are collectively reflected in the expansion of Microsoft Cloud's overall revenue. The initial phase of investment is turning into results, reflected in specific data: Microsoft Cloud revenue reached $59.3 billion, a year-over-year increase of 27%, with Azure and other cloud services revenue growing 43% year-over-year. Additionally: Microsoft's commercial remaining performance obligations grew 84% to $678 billion; Microsoft 365 Copilot has over 30 million paid seats. This can be understood as a large number of confirmed paid orders; AI has been integrated into Microsoft's existing business model, effectively upgrading the narrative. Capital expenditures themselves are not the target of market punishment, after all, Microsoft is also a major investor in AI infrastructure. But the premise is that the company must show the market that these investments have reached the consumer end, that capital expenditures ultimately translate into products, customers, and revenue, making them assets. If they remain only at the construction stage, then they are a cost. Of course, other companies also have investments, but many companies at the forefront have already been anticipated by the market. Meta's Q2 revenue was about $60.8 billion, a year-over-year increase of about 28%, with advertising revenue also maintaining about 27% growth. This means the core business has not deteriorated. But the market's attention in the current environment is no longer on the core business. Meta's operating cash flow for the quarter reached about $31.86 billion, a year-over-year increase of about 25%; meanwhile, capital expenditures surged to about $31.08 billion, a year-over-year increase of about 83%. Subtracting the two, free cash flow is only $784 million, down about 91% from $8.55 billion in the same period last year. Revenue is growing, operating cash flow is growing, but the ultimately available free cash is almost entirely consumed by capital expenditures. I believe this is the divergence in Meta's earnings report this time. However, this does not mean Meta's strategy is a failure (although it has made mistakes in the past). I still attribute it to the overall market sentiment impact; now sentiment has reached an extreme, and the market tends to overinterpret any unfavorable factors in AI concept stocks. Especially for AI concept stocks of this scale, valuations already include some future expectations. So when the actual realization speed is lower than market expectations or similar benchmarks, even if fundamentals have not obviously deteriorated, stock prices will reflect this expectation gap in advance. Therefore, if we look at the two earnings reports together, neither Microsoft nor Meta has reduced AI investment; to some extent, both companies are accelerating. From the data, my above point is that cash flow data affected the short-term market outcome. But looking deeper, it is because they are at different commercialization stages; they are not purely infrastructure companies but lean more towards [AI business model empowerment], expanding on existing stock. Similar logic presents different stages to the market; both are betting, Microsoft has already begun to reap incremental gains under the AI-empowered model, but Meta's results are not yet obvious and cannot be presented as an independently verifiable model. Having said this, I think my core viewpoint is relatively clear: Meta's short-term market decline, I prefer to attribute it to excessive emotional overinterpretation, not a major change in fundamentals. At the same time, Microsoft's relative strength over Meta is also because AI revenue has entered the acceptance phase earlier. But Meta, due to its own user base, even has an advantage compared to Microsoft, will also enter the acceptance phase in the future. In the long term, platforms with users, data, and distribution capabilities will still be the biggest beneficiaries of AI. So the current poor macro sentiment in the sector, while leveraging these targets that inherently have consumer endpoints, also provides the market with an opportunity for expectation gaps. Because the market is currently trading on short-term validation, not long-term value.