--- title: "A $148 billion bet: Why does Goldman Sachs believe the market has misread Microsoft's capital expenditures?" type: "News" locale: "en" url: "https://longbridge.com/en/news/275906352.md" description: "Goldman Sachs' latest report believes that short-term pessimism is overshadowing the rationality of Microsoft's mid-term strategy. The company is proactively shifting computing power from Azure towards internal research and development as well as Copilot, with a more balanced allocation of new GPUs. This strategic delay sacrifices short-term growth but reserves space for long-term moats and monetization. Copilot has already begun to make a substantial contribution to growth, proving that Microsoft's strategic advantage as a \"fast follower\" is becoming evident. The current decline reflects more of an emotional misalignment rather than a fundamental turning point" datetime: "2026-02-13T13:19:20.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/275906352.md) - [en](https://longbridge.com/en/news/275906352.md) - [zh-HK](https://longbridge.com/zh-HK/news/275906352.md) --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/275906352.md) | [繁體中文](https://longbridge.com/zh-HK/news/275906352.md) # A $148 billion bet: Why does Goldman Sachs believe the market has misread Microsoft's capital expenditures? Since the release of the financial report on January 28, Microsoft's stock price has fallen by about 15%, with market concerns focusing on two core issues: first, capital expenditures have been significantly revised upward, but the growth guidance for Azure has not been raised in tandem, raising questions about the logic of return on investment; second, the Office 365 business faces competitive threats from a new generation of AI tools. Goldman Sachs stated in its latest report that **short-term market concerns are overshadowing the rationality of Microsoft's mid-term strategy.** According to the Wind Trading Desk, the report points out that **Microsoft is proactively adjusting its computing power allocation structure**: the proportion allocated to internal research and development and Copilot has increased from about 10% to 20%. If all new GPU capacity is allocated to Azure, its growth rate in the second quarter could reach nearly 40%, **which means some growth potential has been strategically deferred.** In terms of capital expenditure, it is expected to reach $148 billion in fiscal year 2026, a 400% increase from fiscal year 2022. The current computing power allocation is approximately 70% for Azure and 30% for Copilot and internal research and development; the allocation of new GPUs is becoming more balanced, approaching 60/40. Goldman Sachs believes that **Microsoft is prioritizing short-term Azure revenue for long-term strategic layout, which will translate into a stronger moat and returns in the mid-term.** ## "Iceberg Model": Visible and Invisible Computing Power Investment Goldman Sachs introduced the "iceberg analogy" framework in its latest report to interpret Microsoft's current computing capital expenditure strategy. The so-called "above water" refers to computing power investments that are directly monetized and reflected in Azure and Office 365 quarterly performance; **while "below water" refers to computing deployments that have not yet generated direct revenue but are crucial to Microsoft's long-term strategy. This part of the investment, although not reflected in short-term reports, reserves space for future monetization.** Microsoft's management stated that if all new computing capacity is allocated to Azure, its growth rate in the second quarter, calculated at constant exchange rates, will exceed 40%. The actual reported growth rate is 38%, and Goldman Sachs estimates that if fully allocated, Azure's growth rate could reach the low 40% range, equivalent to an increase of about 5 percentage points. This confirms that Microsoft is strategically deferring some computing power for internal research and development and long-term layouts such as Copilot. ## Computing Power Allocation Update: Significant Increase in Internal Use Proportion **Goldman Sachs has raised the proportion of computing power used internally by Microsoft in its latest report from about 10% to 20%.** This adjustment is based on guidance from the second quarter earnings call and the latest tracking of first-party applications such as Copilot and the usage of top engineers' tokens. Microsoft has previously stated its commitment to developing cutting-edge AI models. Regarding the use of computing capital expenditures in fiscal year 2026, Goldman Sachs has established a detailed framework: - **Azure AI computing power** occupies the largest share, applicable to a 4-5 times "computing power/quarterly new revenue" multiplier, meaning that every $1 of capital expenditure can be converted into $0.20 to $0.25 of revenue in the next quarter This section implies that the gross margin over a six-year lifecycle is lower than the current core level of Azure. - **Internal R&D computing power** covers employee AI adoption, increased developer computing power demand, and the development of internal models for Microsoft AI (MAI). - **Application computing power** includes pay-per-use AI computing power and unmonetized Copilot computing power (such as Bing and free Copilot applications), applicable to a 3-4 times multiplier. Overall allocation, **Goldman Sachs estimates that about 70% of total computing power in fiscal year 2026 will be directed towards Azure, with 30% used for Copilot and internal purposes. In the allocation of new GPUs, the ratio is more balanced, approaching 60/40.** ## Four Major Driving Factors for Increased Internal Allocation Microsoft's computing power demand is being driven by four structural factors, leading to a profound evolution in capital expenditure structure. First, **the adoption rate of Copilot continues to rise.** Both Microsoft internal employees and external customers are increasingly reliant on Copilot, directly driving computing power consumption. Second, **internal R&D efforts have significantly increased.** Microsoft stated in February that it is developing cutting-edge AI models. With the IP access agreement with OpenAI until 2032, the company will build dedicated models based on branch versions of OpenAI models tailored to its domain data, focusing on enterprise-level Copilot applications. Third, there is a fundamental shift in the operating expenditure structure. **Microsoft expects future operating expenditure growth to be driven more by capital expenditure rather than employee numbers** due to the increasing computing power intensity of engineering workloads. Employees using internal Copilot will also require incremental computing power support. This trend has already been reflected in the 5% layoffs in 2023 and multiple rounds of personnel adjustments in 2025. Fourth, **the overall baseline computing power level of the software stack is rising.** Customers have formed basic expectations for AI features embedded in existing applications and are no longer accepting separate charging models. For large vertically integrated companies like Microsoft, this change directly translates to increased capital expenditure on computing power; for other software companies, it converts to token fees paid to third-party LLMs or large-scale cloud service providers. ## Path to Improving Azure Market Sentiment In the coming quarters, as new capacity comes online in phases, the trade-off between Azure growth and first-party applications and internal R&D computing power will no longer be a zero-sum game. **This means Microsoft can advance Copilot and cutting-edge model development without sacrificing Azure's growth rate.** **The core obstacle facing current investors is insufficient information disclosure.** Against the backdrop of continuously revised capital expenditure intensity and fluctuating Azure expectations, the market urgently needs an analytical framework that distinguishes between Azure and other business capital expenditures, thereby establishing a clearer judgment on medium- to long-term growth and investment returns. Goldman Sachs estimates that Microsoft had about $4 billion in GPU expenditures in the second quarter that were not directly related to Azure or monetized Copilot business. If this portion of computing power is invested in Azure and a 5 times "computing power/quarter new revenue" multiplier is used (implying a three-year payback period), Azure could gain about 3% upside potential, corresponding to a 5 percentage point increase in growth rate, raising the second quarter report growth rate to 44% (43% when calculated at fixed exchange rates) Compared to other hyperscale cloud service providers, Microsoft relies more on internally developed software solutions, which to some extent occupies capacity that could have been allocated to Azure. For example, the computing power demand arising from the increased adoption of Microsoft's AI coding tools is counted as capital expenditure, while other vendors that purchase third-party software reflect this as operating expenditure. This structural difference also affects the market's assessment of Azure's growth potential. Recently, some funds have flowed to Google, primarily based on three factors: Gemini's competitive positioning, the potential opportunities of TPU, and a relatively small exposure to application businesses. Notably, after Google announced its Q4 2025 results on February 4, its stock price fell, as its 2026 capital expenditure was about 50% higher than market expectations. However, the conference call provided a clear disclosure framework—60% of capital expenditure is allocated to servers, with slightly more than half of the computing power assigned to cloud business. Goldman Sachs believes that if Microsoft can provide similar dimensions of disclosure, it will help rebuild investor confidence. ## Strategic Rationality of Copilot Investment **Microsoft is positioning Copilot as an enterprise-level LLM solution that balances performance and cost through a differentiated delivery mechanism.** Copilot 365 is priced at approximately $30 per user per month and integrates multiple models; in contrast, ChatGPT Enterprise is priced at about $60, offering unlimited access to GPT-4. Individual users may still prefer the simplicity of the latter, but enterprises need to weigh performance against cost when making decisions and assess the long-term strategic value brought by the abstraction layer. In terms of contextual intelligence, WorkIQ is becoming an important vehicle for Microsoft to showcase its differentiated advantages. This tool was launched at the Ignite conference last November and reveals internal collaboration patterns and work trends within enterprises by analyzing anonymized, aggregated Microsoft 365 data, providing data support for embedding AI into business processes. Microsoft also demonstrates a unique strategic advantage as a "fast follower." Leveraging strong distribution channels, enterprise-level security capabilities, and service support, Microsoft can quickly introduce consumer-validated technologies into enterprise scenarios. The past three years of practice have shown that the iteration cycle of enterprise software stacks is long, while management's urgent demand for AI implementation is prompting enterprises to seek incremental value beyond their existing software stacks, which is precisely the opportunity for Microsoft to enter. **More importantly, Copilot has begun to make a substantial contribution to the growth of M365 commercial cloud revenue.** This business is expected to reach $89 billion in fiscal year 2026. Excluding the impact of one-time items, M365 commercial cloud has maintained stable revenue growth over the past three quarters, reversing the previously ongoing slowdown. Goldman Sachs believes that the monetization of Copilot is effectively offsetting the pressure from the E5 upgrade cycle nearing its end and the addition of low ARPU groups (such as regional markets, small and medium-sized enterprises, and frontline employees). The ARPU growth rate over the past three quarters has risen to 10-11%, significantly higher than the previous year's 5-7%, which also indirectly confirms that the adoption of Copilot is accelerating ### Related Stocks - [Microsoft Corporation (MSFT.US)](https://longbridge.com/en/quote/MSFT.US.md) - [YieldMax MSFT Option Income Strategy ETF (MSFO.US)](https://longbridge.com/en/quote/MSFO.US.md) - [Direxion Daily MSFT Bear 1X ETF (MSFD.US)](https://longbridge.com/en/quote/MSFD.US.md) - [T-Rex 2X Long Microsoft Daily Target ETF (MSFX.US)](https://longbridge.com/en/quote/MSFX.US.md) - [Direxion Daily MSFT Bull 2X Shares (MSFU.US)](https://longbridge.com/en/quote/MSFU.US.md) - [GraniteShares 2x Long MSFT Daily ETF (MSFL.US)](https://longbridge.com/en/quote/MSFL.US.md) ## Related News & Research - [Microsoft Corporation $MSFT Shares Sold by Cullen Capital Management LLC](https://longbridge.com/en/news/277905328.md) - [Microsoft Copilot to hijack your browser... for your own convenience](https://longbridge.com/en/news/277944537.md) - [Microsoft reportedly eyes E7 tier to make AI agents pay their way - like the humans they'll replace](https://longbridge.com/en/news/277622185.md) - [Microsoft-Backed OpenAI Reportedly Eyeing Contract With NATO](https://longbridge.com/en/news/277776619.md) - [Microsoft’s Copilot Tasks AI uses its own computer to get things done](https://longbridge.com/en/news/277112228.md)