To The Moon

At a time of mixed long and short signals, why does the market place more on $Microsoft(MSFT)$ "short signals" after earnings?

The following this reason can not be ignored

1. Cloud revenues falling short of expectations is the original sin; previously expectations were given too high and actually not met, a failure of expectation management;

2. Microsoft's Capex investments rely heavily on cloud revenues to be realized, and once revenues fall short of expectations, it may make investors doubt the efficiency of its Capex investments

   

Earnings and Market Feedback

MSFT fell within minutes of releasing its 25Q2 earnings report, then the decline narrowed generally to just 1% at most, but then fell over 6% on the call.This kind of volatility is not common for Microsoft and reflects market concerns about its earnings performance.

• While the overall performance exceeded expectations ($69.6B +12% YoY), the cloud growth rate Missed, with Azure growing at a faster rate (+31% vs 32% Consensus), and total cloud-related revenues came in below expectations ($40.9B vs. $41.1B).Earlier management mentioned in the call that Azure would pick up speed in the second half of the fiscal year, which it actually did not do;

• The impact of exchange rate shocks (the stronger dollar in CY24Q4 was more of a surprise) FX gains and losses would have reduced revenue growth by two percentage points across divisions;

• Large Capex beat expectations in the quarter and capex is expected to continue to increase over the next few years.

  

Investment Highlights

1. Cloud computing revenue fell short of expectations.

   1. Azure grew 31% at constant currency, less than the lower end of the previous guidance range of 31% to 32%, and because the company's early guidance was that H2 would regain its growth rate, buyers were more optimistic in their expectations for growth in the quarter, some greater than 32%;

   2. Meanwhile, total cloud-related revenue for Q2 came in at $40.9 billion, below expectations of $41.1 billion.

   3. The company revealed that 13pc of Azure's growth came from contributions from AI offerings, above analysts' expectations of 12.2pc

2. Eye-catching signals: new commercial contracts, productivity tools and LinkedIn business expected to accelerate.

   1. Productivity and Business Streams segment revenue, which includes Office software such as Microsoft 365 Copilot AI tools, was $29.4 billion, +14% y/y, up from +11% in Q3;

   2. AI-related revenue exceeded $13 billion on an annualized basis in the quarter, up 175% year-over-year.

   3. Dynamics 365 revenue grew 15% in the same timeframe, and sales of its enterprise-oriented social network, LinkedIn, grew 9%;

   4. The value of new commercial contracts signed was +67% year-over-year according to the company, higher than historical levels;

3. Substantial growth in Capex

   1. Capex reached a record high of $22.6 billion in the quarter, up 96% year-over-year;

   2. Expect to continue to increase capex in the upcoming fiscal years, with approximately $80 billion invested in Q3 FY25, most of which will be spent on AI-related infrastructure

   3. Flexible pools of compute resources are being built to balance training and inference, with global placement and software optimization to reduce inference costs to drive demand.

   4. At the same time, there will be no one-time over-purchases given Moore's Law and the uplift from optimization.

   5. Operating cash flow of $22.3 billion (+18% YoY) and free cash flow of $6.5 billion (-29% YoY).

Attitude toward DeepSeek

1. Microsoft incorporates DeepSeek's R1 model into Azure AI Foundry and GitHub

2. CEO Nadella sees DeepSeek as good news for Microsoft: as the cost of AI goes down, and the price of inference computation goes down, people will spend more, more apps will be developed, and AI will become more pervasive, both in terms of saving computing power and still being open source.

   1. More people are buying servers, except it's called the cloud.So when the price of Token goes down, so does the price of inferential computing, which means that people can consume more and more applications will be written.Interestingly, when I mention these fairly powerful models, it's hard to imagine that we could actually run a model that requires a fairly large cloud infrastructure on a personal computer in the early part of the decade and a half.This optimization means that AI is going to become much more common.So for hyperscale cloud providers like us, and PC platform providers like us, this is good news for me.

3. AI development is following Scaling's Law, with increasing efficiency in training and reasoning, and cost-effectiveness of hardware and models.

Why did the market sell off twice in after-hours?

1. Cloud growth rate is less than expected, especially less than previous guidance, the market is expecting a lot, is a failure of expectation management;

2. Increased capex with less-than-expected cloud revenue growth and the market will doubt your ability to commercialize;

3. Unlike META, which can be reflected in advertising revenue very quickly, Microsoft needs to be reflected in Azure services, and compared to META's competitiveness in the social media space, Azure's competitive position in cloud services is much less favorable