Microsoft and Meta Release Earnings Reports on the Same Day
On July 29, Microsoft and Meta announced their earnings reports on the same day, delivering two AI performance reports with completely opposite trends.
Microsoft's stock price once rose by more than 8% after hours. Right after wrapping up the earnings call, Microsoft CEO Satya Nadella posted on X that "This is an exceptionally strong finish to a record fiscal year for Microsoft, and even greater opportunities lie ahead."
It is reported that Microsoft's revenue for fiscal year 2026 reached $331 billion, up 18% year over year; Microsoft Cloud hit $214 billion, up 27%; Azure hit $100 billion, up 41%.
Note: Microsoft's fiscal year runs from July 1 of each year to June 30 of the following year. Therefore: July to September 2025 is the first quarter of Microsoft's 2026 fiscal year; October to December 2025 is the second quarter of Microsoft's fiscal year; January to March 2026 is the third quarter; April to June 2026 is the fourth quarter of the fiscal year.
On the other side, another tech giant Meta also released its earnings news. Meta's revenue in the second quarter reached $60.8 billion, up 28% year over year, making it one of the fastest-growing quarters in recent years.
The daily active users of its portfolio of apps reached 3.6 billion, up 3% year over year; ad impressions rose 14%, and the average ad price increased by 12%. Judging from its core social network and advertising business, Meta remains a powerful cash cow.
However, noticeable pressure has emerged in its income statement and cash flow statement.
Meta's total costs and expenses for the quarter reached $42 billion, up 55% year over year; operating profit dropped 8% to $18.78 billion, with the operating margin falling from 43% a year earlier to 31%; net profit decreased 14% to $15.85 billion, and earnings per share came in at $6.18, which fell below market expectations.
Meta's stock price once dropped by 10% afterwards.
With earnings reports released on the same day, why did the market show completely different attitudes towards Microsoft and Meta?
The answer lies in the two companies' respective AI investments and layouts.
Microsoft's Confidence: AI Is No Longer Just an Expense
Microsoft's revenue in the fourth quarter of fiscal year 2026 reached $90 billion, up 18% year over year; operating profit was $40.6 billion, and net profit was $35.8 billion, up 31% year over year.
It should be noted that the net profit includes a $3.2 billion gain related to its investment in Anthropic. But after excluding investments and one-off items, Microsoft's adjusted net profit still increased by 22% year over year, with revenue, profit and earnings per share all exceeding the company's previous guidance.
What truly excited the market was not the total revenue, but Microsoft's first disclosure that Azure's full-year revenue had exceeded $100 billion.
In this quarter, revenue from Azure and other cloud services rose 43%, beating market expectations; Microsoft also forecasts that Azure will grow by approximately 45% at constant currencies in the next quarter. Microsoft Cloud's revenue for the quarter reached $59.3 billion, up 27% year over year, and its remaining commercial performance obligations rose to $6.78 trillion, up 84% year over year.
More critically, Microsoft stated that the approximately $500 billion in new contract commitments added this quarter all came from customers outside the top US foundational model companies.
This set of data answers the question that Microsoft has been repeatedly asked over the past few quarters: How much of Azure's growth comes from a small number of big customers such as OpenAI, and are enterprises really willing to pay for AI?
Microsoft's answer is that demand is spreading from model companies to a wider range of enterprise customers. Azure is no longer only providing training computing power for companies like OpenAI, but also hosting enterprise databases, Agent platforms, inference services and various types of models.
The number of paid seats for Microsoft 365 Copilot also grew from 20 million in the previous quarter to more than 30 million, higher than the 26.9 million expected by analysts. Microsoft Foundry has reached 100,000 customers, with revenue more than doubling year over year; two months after the launch of Agent 365, nearly 40 million Agents have been registered across tens of thousands of enterprises.
The significance of these figures is that Microsoft is now able to charge on both the AI infrastructure and AI application ends: at the underlying layer, it sells computing, storage, database and model services through Azure, and at the upper layer, it charges subscription fees through Copilot, GitHub, Microsoft 365 and security products.
Nadella's wording in the earnings report has also changed. He no longer simply emphasizes how powerful the model is, but proposes to promote the "cost-to-outcome curve" to enable customers to convert Tokens into business results.
This statement points exactly to what Microsoft most needs to prove right now: AI is not just about generating Tokens, but about integrating into enterprise processes, ultimately generating revenue, saving manpower or improving efficiency.
With $41 billion in capital expenditure, why did Microsoft's stock price rise instead?
Microsoft's capital expenditure for the quarter reached $41 billion, up more than 70% year over year, while free cash flow fell 23% year over year to $19.6 billion. Judging only from the scale of investment, Microsoft is no more conservative than Meta. In the next quarter, Microsoft expects capital expenditure to reach approximately $50 billion.
Moreover, Microsoft adjusted its 2026 capital expenditure forecast from the previous approximately $190 billion to $175 billion, which is mainly related to changes in the accounting treatment of long-term data center leases. The company extended the amortization period for some data center leases from 15 years to 25 years, and its actual construction plans have not been reduced as a result.
The outstanding data center lease commitments that Microsoft has not yet executed have reached $3.291 trillion, and the relevant leases will be launched successively from fiscal year 2027 to 2033.
In other words, Microsoft has not suddenly hit the brakes on AI investment.
The market still chose to reward Microsoft because the new computing power can be quickly converted into revenue.
Microsoft CFO Amy Hood said that demand for Azure still exceeds existing supply. When the engineering team improves the utilization efficiency of CPU and GPU clusters, or shortens the server launch time, the newly released extra capacity will be quickly bought by customers within the same quarter.
This forms a relatively investor-friendly closed loop: insufficient computing power means customer demand already exists; new data center investment is not about building first and then looking for customers, but expanding capacity when a backlog of orders and contracts has already formed; once servers go online, they can quickly generate revenue through Azure, model hosting and enterprise software.
Capital expenditure remains huge, but Microsoft has been demonstrating how each new batch of GPUs can be reflected in the income statement.
Meta's Problem Is Not Lack of Growth, But Collapsing Cash Flow First
In contrast, Meta's earnings report cannot be simply summed up as "poor business performance".
Not all the profit decline can be attributed to AI.
Meta recognized $2.4 billion in legal litigation costs this quarter, and recorded $1.18 billion in severance costs due to layoffs in May. CFO Susan Lee said that if legal and severance expenses were excluded, Meta's operating profit should have increased by 9% year over year instead of decreasing by 8%.
What really unsettled investors is free cash flow.
Meta's cash flow generated from operating activities for the quarter was $31.86 billion, but capital expenditure reached $31.08 billion, leaving only $784 million in free cash flow in the end, down 91% from $8.55 billion a year earlier.
In the first quarter, Meta's free cash flow was still $12.39 billion. That means within one quarter, new data center, server and infrastructure expenditures have almost eaten up all of Meta's operating cash flow.
Meta also narrowed its 2026 capital expenditure range to $130 billion to $145 billion, raising the lower limit from $125 billion to $130 billion. At the beginning of this year, Meta's forecast was only $115 billion to $135 billion.
Calculated at the upper limit, its annual capital expenditure is approximately twice that of 2025.
Both are AI investments, why is it harder for Meta to justify its returns?
Microsoft sells computing power and enterprise software, while Meta's most mature business model to date is still advertising.
AI is certainly already helping Meta make money.
Recommendation algorithms can increase user stay time, advertising models can improve matching and conversion, and generative AI tools can also lower the threshold for advertisers to create materials. The simultaneous growth of ad impressions and unit price shows that these investments are already supporting its core business.
Microsoft can clearly tell investors how much Azure has grown, how many Copilot seats have been sold, and how many contracts customers have signed. It is difficult for Meta to separate how much of its advertising revenue can be directly attributed to a specific large model, a specific data center or a specific AI product.
The bigger problem is that the computing power Meta is building now is not only for optimizing ads.
Mark Zuckerberg stated on the earnings call that a large amount of computing resources will be used to train models, boost core businesses, deliver personal agents and new products. Meanwhile, Meta also plans to develop businesses targeting large enterprise customers.
He regards personal AI agents as a potentially huge consumer market, and hopes to further expand into fields such as APIs, enterprise productivity tools and business agents. He even published an article titled *The Future of AI Belongs to Everyone* in *The Wall Street Journal* to show his confidence in the future of personal agents.
This means Meta is attempting to expand from an advertising platform to a more full-fledged AI platform company.
But in the enterprise market, Meta is facing Microsoft, Google and Amazon.
These companies already have cloud platforms, enterprise customer relationships, identity authentication, databases, office software and sales networks. Meta has a huge consumer user base, but lacks the mature enterprise distribution entry that Microsoft has.
Therefore, what Meta is paying now is the transformation cost: on one hand, it needs to continue improving the advertising efficiency of Facebook, Instagram, WhatsApp and Threads, and on the other hand, it needs to train foundational models, build personal agents, and try to open up revenue from enterprise AI and cloud services.
Microsoft is expanding lanes on a road that is already generating revenue; Meta is building a new road, and is still not entirely sure who to charge in the end and what charging model to adopt.
Microsoft Wins, Meta Loses?
These two earnings reports do not mean that Microsoft has completely solved the AI commercialization problem, nor do they mean that Meta's AI strategy has failed.
Microsoft's capital expenditure is still growing, and free cash flow is also declining. More than $3 trillion in future data center lease commitments means it has bet on a multi-year heavy asset plan. Its Windows, devices and Xbox businesses also declined respectively this quarter.
Meta's core advertising business is still growing rapidly. Excluding legal and layoff expenses, its operating performance is not as bad as the earnings report appears. Meta has 3.6 billion daily active users, and any small improvement in recommendation or advertising efficiency can be converted into considerable revenue.
What the market is actually voting for is the visibility of returns.
Microsoft can already present Azure revenue, Copilot seat numbers, Foundry customers, contract backlog and free cash flow to explain how AI investments translate into current-period revenue, while what Meta shows is mostly ad performance improvements, as well as future opportunities in personal agents and enterprise AI.
This is also an increasingly clear dividing line in the current AI capital expenditure competition: investors are not opposed to companies spending money, but they require companies to prove when new GPUs will generate revenue, whether customers are willing to pay continuously, and whether cash flow can cover the next round of expansion.
Now, Meta needs to give the outside world a clear answer: beyond the advertising business, what products will the huge computing power it has built eventually turn into, and who will pay for them.
References:
https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q4
https://finance.yahoo.com/technology/ai/articles/meta-microsoft-report-ballooning-ai-203954295.html
https://x.com/juan_gonzalesP/status/2082603552975761622
https://www.wsj.com/opinion/the-ai-future-is-for-everyone-a0c24e20?mod=hp_opin_pos_2
This article is from the WeChat official account "AI Front", written by Dong Mei, published with authorization from 36Kr.