--- title: "Microsoft Research Special Issue — Why AI Makes Microsoft Stronger, Not Weaker" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/43066283.md" description: "One-sentence summary: AI will not replace $Microsoft(MSFT.US)'s software; AI will make Microsoft's software more valuable. This is the biggest point of divergence for many investors. Why was the market initially bearish on Microsoft? From 2023 to 2025, although Microsoft's stock price has risen in the long term, a question repeatedly arose in the market: When will Microsoft start making money from its investment of tens of billions of dollars annually in building AI? Investors are worried about four things: First, capital expenditures (CapEx) are getting higher and higher; Second, GPUs are getting more expensive; Third, AI inference costs are very high; Fourth..." datetime: "2026-07-30T08:48:16.000Z" locales: - [en](https://longbridge.com/en/topics/43066283.md) - [zh-CN](https://longbridge.com/zh-CN/topics/43066283.md) - [zh-HK](https://longbridge.com/zh-HK/topics/43066283.md) author: "[长期主义接班人](https://longbridge.com/en/profiles/16386327.md)" --- # Microsoft Research Special Issue — Why AI Makes Microsoft Stronger, Not Weaker **Summary in one sentence: AI will not replace**$Microsoft(MSFT.US) **'s software; AI will make Microsoft's software more valuable.** This is the biggest point of divergence for many investors. **Why was the market initially bearish on Microsoft?** From 2023 to 2025, although Microsoft's stock price rose over the long term, the market repeatedly raised a question: **When will Microsoft start making money after investing hundreds of billions of dollars in building AI every year?** Investors were worried about four things: First, capital expenditures (CapEx) are getting higher and higher; Second, GPUs are getting more expensive; Third, AI inference costs are very high; Fourth, Copilot might not sell well. So the market worried whether Microsoft had turned into a "money-burning company." This was an important factor suppressing valuations at the time. **What the market overlooked** Many people think AI is just ChatGPT. In reality, what Microsoft truly sells is not models, but **AI infrastructure.** Suppose a company wants to deploy AI. What does it need? It's not just having one model. It needs: GPU → Cloud Computing → Database → Identity Authentication → Security → Knowledge Base → Office Systems → Code Platform → Agent → AI Applications. Microsoft has almost all of these. Therefore, Microsoft's revenue sources extend far beyond just models. **Azure is the biggest winner in the AI era** Many people focus on OpenAI, Gemini, and Claude. Actually, the one truly making money might be Azure. Because all AI requires training, inference, storage, databases, networking, and security, all of which run on the cloud. The more AI calls in the future, the higher Azure's revenue. So what Microsoft is truly betting on is **the "highway" of the AI era,** rather than any single car on that highway. **What does Copilot really sell?** Many people think Copilot is just an AI chatbot. That is completely incorrect. What Copilot really sells is **efficiency.** For example. Previously, it took employees three hours to write a report. With Copilot, it might take only one hour. If an enterprise has 10,000 employees, saving hundreds of thousands of hours a year, then enterprises would be willing to pay subscription fees. Microsoft is not selling AI. Microsoft is selling **enterprise ROI (Return on Investment).** **Why is Microsoft easier to monetize than many other AI companies?** This is an interesting question. An AI startup first needs to find customers. Microsoft doesn't have to. Microsoft already owns millions of enterprise customers. For instance: Enterprises have already purchased Microsoft 365. Microsoft only needs to add Copilot, and many customers will try it directly. Therefore, Microsoft's cost of acquiring AI customers is far lower than that of many startups. **GitHub Copilot—the product that first validated the AI business model** GitHub Copilot is one of Microsoft's earliest AI products to achieve large-scale commercialization. Because programmers write code every day, AI can directly help with auto-completion, function generation, bug fixing, code explanation, and improving development efficiency. Programmers use it almost every day. Therefore, GitHub Copilot proves one thing: **Enterprises are willing to continuously pay for AI that truly improves efficiency.** This is closer to Microsoft's long-term business model than simply pursuing chat capabilities. **In the Agent era, Microsoft will become stronger** The biggest change in the next few years may not be chatbots, but AI Agents. What is an Agent? For example: An employee says, "Organize all customer data." The Agent automatically opens Outlook → reads emails → opens Excel → queries CRM → generates PPT → sends via Teams. The entire process is completed automatically. Many of these software tools are owned by Microsoft itself. Therefore, the more popular Agents become, the higher the value of the Microsoft ecosystem. **Microsoft's true AI strategy** Many investors believe Microsoft's greatest success is its investment in OpenAI. That is not the case. What makes Microsoft truly formidable is integrating AI into Azure, Office, GitHub, Windows, Security, Dynamics → across the entire Microsoft ecosystem. Microsoft did not create AI as an isolated product but transformed AI into a capability for all its products. This approach is very similar to Apple's strategy back then: AI is not a product; AI is a feature. **Will AI kill Office?** Many people worry that in the future, when everyone asks AI directly, why would they still need Word? My view is exactly the opposite. In the future, content generated by AI will still ultimately require: documents, spreadsheets, presentations, emails, meetings, approvals, and corporate knowledge. These all need to land within specific workflows. Therefore, Office will not disappear; instead, it will become smarter. In the future, Office may not just be editing software but will become the entry point for AI-assisted work completion. **Microsoft's greatest advantage is not its AI models** Many investors compare OpenAI, Google, Anthropic, and xAI daily to see who has the best model, but I am more concerned with whether Microsoft has the ability to turn AI into revenue. I believe Microsoft's true lead lies in its customers, channels, data, platform, cloud computing, and enterprise relationships. These are more important than the models themselves because model leadership may last only a few months, while an enterprise ecosystem can endure for decades. **When will the AI investment finally break even?** This is currently the most pressing question for the market. The answer is: No one knows the exact time. However, we can observe several signals: First: Growth in Azure AI revenue. Second: Growth in paid Copilot users. Third: Number of enterprise Agents. Fourth: Pricing power for Office driven by AI. Fifth: Whether AI improves overall gross margins. As long as these indicators continue to improve, Microsoft's hundreds of billions in CapEx will eventually gradually turn into revenue. Of course, this process may take years, and there may also be demand fluctuations and intensified competition during this period. There is no inevitable or linear return path. **My investment notes** After researching Microsoft, this is an important viewpoint I have formed. Many investors view Microsoft as: **an AI company,** but I do not fully agree. I prefer to define Microsoft as: **a platform company that uses AI to strengthen its entire enterprise digital ecosystem.** AI itself will become increasingly ubiquitous, and the gap between models may gradually narrow. But what is truly hard to replicate is: - Decades of enterprise customer relationships; - The Azure cloud platform; - The Microsoft 365 office ecosystem; - The GitHub developer ecosystem; - Enterprise identity, security, and management systems. AI makes the connections between these businesses tighter and allows Microsoft to more easily increase the long-term value of individual customers. Therefore, my investment in Microsoft is not a bet that a specific model will remain leading in the long term, but a bet on: **That when enterprises adopt AI in the future, Microsoft will remain one of the most critical foundational platforms.** ### Related Stocks - [MSFT.US](https://longbridge.com/en/quote/MSFT.US.md) - [002245.CN](https://longbridge.com/en/quote/002245.CN.md) - [TSLA.US](https://longbridge.com/en/quote/TSLA.US.md) - [MSFL.US](https://longbridge.com/en/quote/MSFL.US.md) - [MSFO.US](https://longbridge.com/en/quote/MSFO.US.md) - [MSFD.US](https://longbridge.com/en/quote/MSFD.US.md) - [MSFU.US](https://longbridge.com/en/quote/MSFU.US.md) - [MSFX.US](https://longbridge.com/en/quote/MSFX.US.md) - [MSFY.US](https://longbridge.com/en/quote/MSFY.US.md) - [TSDD.US](https://longbridge.com/en/quote/TSDD.US.md) - [TSLL.US](https://longbridge.com/en/quote/TSLL.US.md) - [TSLQ.US](https://longbridge.com/en/quote/TSLQ.US.md) - [09366.HK](https://longbridge.com/en/quote/09366.HK.md) - [07766.HK](https://longbridge.com/en/quote/07766.HK.md) - [07366.HK](https://longbridge.com/en/quote/07366.HK.md) - [TSLR.US](https://longbridge.com/en/quote/TSLR.US.md) ## Comments (13) - **巨力哥 · 2026-07-30T09:25:53.000Z**: Google's earnings report is also quite good. The only difference from Microsoft is the negative cash flow; I actually think this dip makes it cheaper than Microsoft right now 😂 - **长期主义接班人** (2026-07-30T09:32:12.000Z): Google also bought. Also bought calls. - **FreeBird2025 · 2026-07-30T09:20:25.000Z**: Is seeing 450 as a luxury in this earnings report huge? 😂 - **宁可少赚，不要亏钱** (2026-07-30T09:25:40.000Z): 500 - **长期主义接班人** (2026-07-30T09:43:19.000Z): Let's see how the broader market is doing. Nothing can be said for sure right now. - **螺旋丸爆锤空军 · 2026-07-30T08:55:53.000Z**: ...... didn't hold Microsoft, heavy positions in Tesla + TSMC, got completely stunned - **长期主义接班人** (2026-07-30T09:08:13.000Z): Haha. If you had bought in when I wrote the first one, it would still have been in time. - **螺旋丸爆锤空军** (2026-07-30T09:30:04.000Z): I had some before, sold a bit after making money pre-earnings to buy TSMC, but it got cut in half now, dare not bet on earnings 😂😂 - **长期主义接班人** (2026-07-30T09:42:39.000Z): TSMC's call has been falling continuously, so I keep adding to the position.