Deep Analysis 48 Hours Before Microsoft Earnings: OpenAI Agreement Restructuring Implemented, Three Variables of Azure, Copilot, and Capital Expenditure Determine MSFT Valuation Recovery Path

Microsoft will report FY26 Q3 results after the bell on April 29, 2026, EST. Forty-eight hours prior to the report, on April 27, Microsoft and OpenAI announced a major restructuring of their partnership agreement: Microsoft’s licensing of OpenAI's intellectual property will shift from exclusive to non-exclusive (extended until 2032); simultaneously, Microsoft will cease paying revenue shares to OpenAI, while OpenAI’s revenue share to Microsoft will continue through 2030 with a set cap. This structural change, combined with the upcoming quarterly results, makes this earnings report a critical observation window for assessing Microsoft’s AI transformation path.

This article will dismantle the focus of market controversy around three core data dimensions—Azure cloud business growth, Microsoft 365 Copilot paid seat penetration, and the quality structure of capital expenditure guidance—and provide valuation anchors from a long-term investment perspective based on official Microsoft disclosures and mainstream investment bank consensus.

OpenAI Agreement Restructuring: From Market Panic to Structural Benefit 

The market's biggest short-term concern regarding Microsoft stems from a specific figure. In its FY26 Q2 earnings report, Microsoft disclosed that the company’s commercial backlog had reached $625 billion, a 110% year-on-year increase. However, approximately 45% (about $281 billion) of that came from a single customer: OpenAI. In February 2026, OpenAI cut its compute spending budget for the coming years from approximately $1.4 trillion to $600 billion, leading the market to immediately question the realizability of this portion of the backlog.

The April 27 agreement restructuring provides partial answers to this concern. According to official reports from Microsoft and various sources such as CNBC, Bloomberg, and TechCrunch, the core terms of the new agreement include: First, Microsoft’s intellectual property licensing for OpenAI models and products is extended to 2032 but is no longer exclusive; second, Microsoft will no longer pay any revenue share to OpenAI; third, OpenAI's revenue share to Microsoft (approximately 20%) will continue until 2030, independent of AGI technical progress, but with a total cap; fourth, OpenAI products will still be prioritized for release on Azure, and Microsoft maintains its position as OpenAI's primary cloud partner, although OpenAI can now provide services to any cloud provider including AWS and Google Cloud.

The financial implications of this restructuring should be understood on two levels. In the short term, Microsoft no longer bears the expense of revenue sharing with OpenAI, which is a direct benefit to the gross margins of downstream products like Microsoft 365 Copilot. In the long term, OpenAI's ability to sell services to other cloud providers (including a strategic partnership with AWS worth up to $50 billion) means Microsoft's share of OpenAI's business may drop from 100%, but OpenAI's payments to Microsoft will continue. Under the new agreement, Microsoft effectively "benefits from both ends": it reduces cash outflows while continuing to receive cash inflows, while also mitigating legal risks (particularly potential litigation risks arising from OpenAI’s partnership with AWS) associated with exclusive binding.

More importantly, this agreement restructuring changes the interpretation of the $281 billion OpenAI-related backlog. In an earlier agreement revision in October 2025, OpenAI had already committed to purchasing a total of $250 billion in cloud services on Azure. Even if OpenAI shifts some new workloads to other clouds, the existing Azure contractual obligations do not disappear. Furthermore, because the new agreement removes the uncertainty brought by AGI trigger clauses, OpenAI's payment structure to Microsoft becomes more predictable over the next five years.

The market's immediate reaction confirmed this assessment. On the day the agreement was announced, Microsoft's stock price fell slightly by about 1%, with no significant sell-off. This stands in stark contrast to the "breakup" narrative rendered in media headlines, reflecting that institutional investors' actual interpretation of the agreement terms leaned from neutral to positive.

Azure Growth: The Capacity-Constrained Nature Behind Nine Consecutive Quarters Above 30% 

Azure is currently Microsoft’s most important growth engine and the primary observation metric for this earnings report. According to official data disclosed on Microsoft’s Investor Relations page, Azure and other cloud services achieved 39% year-on-year growth in FY26 Q2 (38% in constant currency). This marks the ninth consecutive quarter that Azure growth has exceeded 30%, ranging from 31% in FY24 Q1 to 39% in FY26 Q2, without a single quarter falling below 30% during that period.

Management's guidance for Azure growth in FY26 Q3 is 37% to 38% (in constant currency). The interpretation of this guidance needs to be built on a three-tier framework. If actual growth stays at 39% or higher, it means AI-driven demand pull and capacity expansion are synchronized, which is the strongest bullish signal; historically, surprises of this magnitude usually correspond to single-day stock price rebounds of 5% to 10%. If growth falls within the 37% to 38% range, meeting company guidance, the market reaction is expected to be primarily consolidation. If it falls below 36%, it will trigger short-term selling, but the underlying root cause would need further deconstruction.

The CFO explicitly pointed out in the FY26 Q2 earnings call that Azure's current growth bottleneck comes from capacity rather than demand. She used the key phrase "capacity受限" (capacity constrained) and stated directly that "customer demand continues to exceed supply." The financial implication of this statement is significant: insufficient demand implies a structural problem where products lack a market; insufficient capacity implies that orders are in place but supply cannot keep up—this is an execution issue and, essentially, a "high-quality problem." Microsoft added nearly 1 gigawatt of data center capacity in Q2 alone, which CEO Satya Nadella listed as one of the most important operational achievements of the quarter during the earnings call.

From an infrastructure perspective, Microsoft’s capacity expansion shows a diversified layout. The newly built Fairwater data center uses a two-story design and liquid cooling technology to support higher GPU density and improve performance and latency for large-scale training. At the chip level, Microsoft is deploying products from NVIDIA and AMD alongside its in-house Maia series accelerators, with the Maia 200 entering service in early 2026. The Maia 200 provides over 10 petaFLOPS of computing power at FP4 precision, offering a total cost of ownership (TCO) improvement of over 30% compared to the latest generation of hardware in Microsoft's existing fleet. This multi-vendor strategy both reduces dependence on a single chipmaker and provides room for optimizing unit compute costs.

Notably, Microsoft itself disclosed that it achieved a 50% throughput increase on one of its highest-throughput workloads (the OpenAI inference tasks driving Copilot). This data has a dual meaning: on one hand, it proves that capacity expansion is translating into actual service capability; on the other hand, it means that even while GPU supply remains tight, software-level optimizations can directly release significant effective capacity.

Therefore, for long-term investors, even if Azure growth temporarily falls below 36% and triggers a market sell-off, as long as the capacity-constrained nature remains unchanged, the decline constitutes a buying window rather than a signal to retreat.

Copilot Paid Seats: 160% YoY Acceleration Behind 3.3% Penetration 

The second key metric is the number of paid Microsoft 365 Copilot seats. This is the core litmus test for Microsoft’s ability to monetize AI.

The FY26 Q2 earnings report officially disclosed Copilot paid seat data for the first time: 15 million seats. Combined with the fact that the total number of Microsoft 365 commercial paid users has exceeded 450 million (specifically about 450 million, up slightly from 446 million in the previous quarter), the current penetration rate of Copilot within its own customer base can be calculated at just 3.33%.

This figure produces diametrically opposed interpretations from different perspectives. From a short-term view, 3.3% penetration is significantly lower than optimistic market expectations for enterprise AI applications. However, from a growth momentum perspective, the number of new Copilot seats in FY26 Q2 grew by more than 160% year-on-year, the largest single-quarter increase ever. At the same time, the number of large enterprise customers deploying more than 35,000 seats tripled year-on-year, with a list of new benchmark customers including Fiserv, ING, NASA, the U.S. Department of the Interior, Westpac, the University of Kentucky, and the University of Manchester. Advertising group Publicis purchased more than 95,000 seats in a single quarter, covering almost all employees. Daily active users of Copilot grew 10x year-on-year, and the number of conversations per user doubled.

This implies two things: first, Copilot is transitioning from an "early experimental deployment" phase to an "enterprise-wide standardized deployment" phase; second, the remaining 96.7% of the Microsoft 365 user base constitutes a massive addressable incremental market for years to come. Every Office customer that moves from zero penetration to paid status will translate directly into high-margin recurring revenue for Microsoft.

For the current quarter (FY26 Q3), the target range for Copilot seats to watch is 20 million to 25 million. If the actual number reaches this range, it means the AI monetization flywheel has entered an acceleration phase; if it remains around 18 million, it represents that penetration is still climbing but in the right direction. It should be noted that after the April 27 agreement restructuring, Microsoft no longer pays a revenue share to OpenAI for Copilot products, which will directly improve Copilot’s operating gross margin. If the seat data in the earnings report is strong, combined with the narrative of improved gross margins, the marginal boost to the stock price will be significantly amplified.

GitHub Copilot is also worth watching. Paid subscriptions reached 4.7 million in FY26 Q2, up 75% year-on-year; the Pro Plus subscription for individual developers grew 77% quarter-on-quarter. Large engineering organizations like Siemens have begun sitewide deployments on the GitHub platform. Penetration speed on the developer side typically leads general office scenarios and can be seen as a leading indicator for the overall expansion of the Copilot ecosystem.

Additionally, Microsoft recently began integrating Anthropic’s Claude models into the Copilot ecosystem. The deeper meaning of this strategic move is to reduce dependence on a single AI model provider (OpenAI). After the new agreement removed the exclusive license, Microsoft can more flexibly combine the capabilities of different model providers to offer differentiated Copilot experiences for various enterprise customer scenarios. The benefits of this architectural adjustment are expected to take two to three quarters to reflect in seat numbers and ARPU metrics.

Capital Expenditure: A Quality Assessment Framework for $110 Billion to $120 Billion 

The third key metric is the amount and structural quality of capital expenditure (Capex). This is also the area where short-side arguments are most concentrated.

According to the company’s FY26 Q2 disclosures, single-quarter Capex reached $37.5 billion, a 66% year-on-year increase and a continued rise from $34.9 billion in Q1. Within this, cash payments for property, plant, and equipment were $29.9 billion, with finance leases accounting for about $6.7 billion (mainly for large data center sites). In terms of structural distribution, the CFO explicitly disclosed: about two-thirds (approximately $25 billion) was for short-lived assets, primarily GPUs and CPUs required for Azure and AI solutions; the remaining one-third (approximately $12.5 billion) was for long-lived assets, namely data center buildings, land, and power infrastructure that can support business operations for more than 15 years. Microsoft's full-year (FY2026) Capex guidance range is $110 billion to $120 billion, a significant upward revision from approximately $80 billion in FY2025.

The market's core concern regarding this scale of Capex is depreciation pressure. Short-lived assets (core computing hardware like GPUs and CPUs) are typically amortized over 3 to 5 years, while long-lived assets are amortized over 15+ years. When Capex nearly doubles within two years, the corresponding depreciation costs will gradually manifest in the coming quarters, putting downward pressure on earnings per share (EPS). The core of the bear case is: if revenue growth from Copilot and Azure AI cannot keep pace with Capex, Microsoft’s operating leverage will turn negative.

In fact, the market's reaction to Capex already gave a clear signal on the day of the Q2 report. FY26 Q2 results were announced after the bell on January 28, 2026; the next day (January 29), Microsoft's stock price plummeted by approximately 10% in a single day, with about $357 billion in market value wiped out—Microsoft's worst single-day performance since March 2020 and the largest single-day dollar-value loss in its history. From the all-time high of about $555 in late October 2025 to a 52-week low of about $356 in early April 2026, Microsoft's stock price corrected by more than 35%. In other words, the bearish scenario for Capex risk has been fully priced in over the past three months.

However, for long-term investors, the total Capex figure is a surface number; what truly determines its investment quality are three structural indicators.

First, the coverage ratio of GPU capacity by contracted customers. In successive earnings calls, Microsoft has emphasized that the vast majority of its new GPU capacity corresponds to specific customer contracts rather than speculative capacity building. The CFO’s explicit statement that "customer demand continues to exceed supply" implies that the "idle risk" for new capacity is low, and depreciation costs can be covered by corresponding service revenue. This is fundamentally different from the Capex strategies of some peers who are purely betting on future demand.

Second, the absolute scale and strategic significance of investment in long-lived assets. While the high proportion of short-lived assets reflects the urgency of current AI inference demand, $12.5 billion per quarter invested in long-lived assets remains a substantial sum. This portion of spending can continue to generate cash flow over a cycle of 15+ years, forming core assets for long-term competitive moats. A key indicator to watch in Q3 is whether the share of long-lived assets remains stable above 30%, as well as changes in the scale of finance leases (primarily corresponding to data center sites).

Third, discipline on the operating expense side. In late April 2026, Microsoft announced its first-ever "voluntary separation program," affecting up to 7% of U.S. employees, or about 8,750 people. Market interpretations of this move are divided. The bear perspective argues that a company with good business prospects does not need to proactively trim headcount, suggesting management concerns about future profitability. However, from a financial management perspective, proactively shrinking labor costs while annual Capex is nearly doubling precisely reflects management's foresight regarding operating leverage: by tightening expenses, they offset the expansionary pressure of depreciation to maintain free cash flow stability.

If Capex guidance is further revised upward in this earnings report but is accompanied by clearer cost control signals (including shrinking labor spending and improved operational efficiency metrics), the overall impact should be viewed as bullish. Conversely, if Capex significantly exceeds expectations and management fails to provide corresponding cost control commitments, one should be wary of short-term stock price pressure.

Valuation Anchors: Historical Position and Upside Potential of 22 to 25 Times Forward P/E 

Integrating the three data metrics into the valuation level, Microsoft currently presents clear relative attractiveness.

As of late April 2026, Microsoft's stock price is approximately $424, corresponding to a forward price-to-earnings (Forward P/E) ratio of about 22 to 25 times. This valuation level is the lowest range of the past three years. For comparison, the S&P 500 currently has a forward P/E of about 21x, and Apple is at about 30x. Among the so-called "Magnificent Seven," Microsoft’s valuation is significantly below average, reflecting that the market has fully priced in concerns over OpenAI risks, Capex pressure, and slowing Azure growth.

Even more historically significant is the comparison with Microsoft’s own valuation mean. Microsoft's average forward P/E over the past five years is 32.9x. If the current valuation merely returns to the five-year mean—without any earnings beats, Azure growth breakthroughs, or an explosion in Copilot penetration—simple valuation recovery alone could bring about 24% upside.

Wall Street analyst consensus further reinforces this view. According to aggregated data from TipRanks, StockAnalysis, MarketBeat, and Benzinga, the median price target for analysts covering Microsoft is approximately $575 to $600, implying 37% to 42% upside relative to the current price. Among all Wall Street analysts covering Microsoft, about 94% give it a "Buy" rating, and not a single one gives a "Sell" rating. Dan Ives of Wedbush Securities set a price target of $625 and named Microsoft his top large-cap AI pick for 2026; Goldman Sachs has a price target of $655, the most optimistic among mainstream investment banks; UBS gives a target of $650; Dan Romanoff of Morningstar maintains a fair value estimate of $600 and lists Microsoft as one of his top picks. The most recent analyst rating (April 23, 2026, by Guggenheim) maintained a "Buy" rating with a price target of $586.

The significance of these numbers is not that "Microsoft will definitely go up," but rather to reveal that the current stock price already reflects relatively pessimistic scenario assumptions. With the valuation at a three-year low, institutional consensus generally bullish, and catalysts (earnings, agreement restructuring) about to land, the negative shock required for a further significant decline far exceeds currently known risk factors.

It is worth adding that Microsoft's cash flow status and shareholder return pace are both at historic highs. FY26 Q2 operating cash flow reached $35.8 billion, a 60% year-on-year increase, primarily benefiting from strong cloud business billing and collections. During the same period, $12.7 billion was returned to shareholders (including dividends and buybacks), a 32% year-on-year increase. This acceleration in capital return reflects the company's commitment to shareholder returns even during a heavy asset investment cycle, serving as another structural factor supporting the valuation.

Furthermore, Microsoft holds an approximately 27% stake in OpenAI (26.79% precisely). When OpenAI completed its Public Benefit Corporation reorganization in October 2025, OpenAI’s overall valuation was approximately $500 billion, corresponding to a book value of about $135 billion for Microsoft’s stake. Microsoft has invested a cumulative total of approximately $13.8 billion in OpenAI since 2019, representing a return of about 10x at current valuations. The value change of this non-operating asset is not fully reflected in Microsoft’s traditional financial metrics, constituting an implicit safety cushion for its valuation.

Core Conclusion from a Long-Term Investment Perspective 

Looking back at Microsoft’s development over the past thirty years, the company has completed three full strategic transformations: from the PC operating system era to the Office productivity software era, and then to the cloud subscription era. Every transition period was accompanied by long periods of stock price consolidation and phases of market skepticism, yet after each transformation was completed, the company's market value jumped to a new order of magnitude.

The fourth transformation is currently underway—the evolution from the cloud era to the AI era. The uniqueness of this transformation is that Microsoft simultaneously controls three key links in the AI value chain: the underlying Azure cloud infrastructure, the middle-layer AI model partnerships (including agreements with OpenAI and Anthropic), and the top-layer Microsoft 365 and GitHub Copilot application portals. This full-stack industry position is unique among the Magnificent Seven and forms the structural basis for a long-term valuation premium.

Regardless of the outcome of the April 29 earnings report, it will not change the basic outline of this long-term narrative. Azure’s data center expansion is a multi-year engineering feat, Copilot’s penetration climb from 3.3% toward 30% or 50% is a multi-year process, and the return on $110 billion in Capex is a multi-year cash flow realization. Short-term stock price volatility reflects the market's reaction to single-quarter data, whereas long-term value accumulation occurs in the continuation of trends behind each quarter's numbers.

For long-term investors, Microsoft is currently in a window with a clear risk premium: valuation is at a three-year low, the long-term narrative framework is intact, and institutional consensus is bullish. This is a critical moment to complete deep due diligence before the earnings report and to update views based on actual data after the report.