How Will Microsoft Corporation (MSFT.US) Realize Returns on Its AI Investments? Morgan Stanley Reviews the Ignite Conference: Flexible Allocation of GPU Compute, Betting on Full-Stack Enterprise AI Commercialization
Morgan Stanley issued a research report reiterating its "Overweight" rating on Microsoft with a price target of US$650.
Morgan Stanley published a research report reiterating its "Overweight" rating on Microsoft Corporation (MSFT.US) with a price target of $650. The Morgan Stanley analyst team led by Adam Wood stated that after Microsoft Corporation held its Ignite conference in San Francisco and held discussions with CEO Nadella and CFO Amy Hood, the firm believes Microsoft Corporation is one of the few software companies that has already realized returns on its AI investments.
Compute Strategy: Not Blindly Maximizing Long-Term GPU Contracts, Retaining a Diversified Compute Services Business
Morgan Stanley noted in the report that on compute resource allocation, Microsoft Corporation management treats GPUs as portfolio assets for allocation and refuses to single-mindedly pursue signing long-term GPU-as-a-Service (GPUaaS) contracts. Management stated that the company will proactively reserve compute resources and supply them simultaneously to its own first-party products, the Foundry platform, and external enterprise customers, rather than locking all compute resources into long-term contracts. From Microsoft Corporation's perspective, model labs are merely important customers and do not represent the entirety of the cloud business; retaining compute flexibility is a proactive investment in the value of its own platform and applications, aimed at capturing monetization opportunities in high-value-added services layered on top of infrastructure.
Regarding the neocloud architecture strategy that the market is focused on, Microsoft Corporation stated that the underlying strategy has not changed, and the construction priority remains self-built data centers, followed by long-term lease models in which Microsoft Corporation holds control over data center fit-out, networking, and hardware equipment. This model allows better control over data center operating efficiency, deployment pace, and equipment capital expenditure. The cancellation of some third-party cooperation orders stems from insufficient service delivery capability on the part of suppliers, and does not represent a shift in Microsoft Corporation's overall compute procurement strategy, nor does it indicate a major change in the market's compute supply-demand landscape.
At the same time, Microsoft Corporation's compute efficiency improvements are not limited to the GPU layer; CPU utilization and storage systems are also key focus areas for efficiency optimization, and newly launched infrastructure inherits this optimization approach, so the efficiency dividend is not confined to the existing server fleet. However, the firm cautioned that efficiency gains cannot be treated as an additional upside variable in compute planning on their own; efficiency improvements can simultaneously raise revenue-carrying capacity and lower delivery costs.
Profitability and Gross Margin: Product-Level Improvement Takes Priority; Cloud Business Mix Will Periodically Weigh on Blended Gross Margin
Morgan Stanley specifically pointed out a key point that the market easily confuses: Microsoft Corporation management's core goal is to improve the product gross margins of Azure and M365 individually; the Microsoft Corporation cloud blended gross margin disclosed in financial reports is merely a result of business development and is not a direct management assessment target.
Even if the profitability of individual products continues to improve, once the share of the high-growth but relatively lower-gross-margin Azure business continues to expand, it will still dilute Microsoft Corporation's consolidated cloud gross margin. Investors cannot simply conclude that the profitability of individual cloud services is weakening just because the blended gross margin is under pressure. When evaluating profitability improvement going forward, focus should be on resource utilization, infrastructure efficiency, model running costs, and the consumption scale of high-value platform services.
Enterprise AI Flywheel: Positioned as a Multi-Model Orchestration Layer, Not Confined to Self-Developed Large Models; Microsoft IQ Closes the Capability Loop
Microsoft Corporation does not bet its competitive position on a single model itself; the core value to enterprise customers comes from the entire system surrounding the model: enterprise business context, model scheduling frameworks, memory capabilities, and security and trust controls. Enterprises can mix and match frontier large models, Microsoft Corporation's self-developed specialized MAI models, and open-weight models; the Microsoft Corporation platform is responsible for selecting the most cost-effective model to complete the work based on task requirements, while leaving control of enterprise business data with the customer and uniformly managing the security, permissions, and cost overhead of all AI agents.
The Microsoft IQ unveiled at this conference integrates Work IQ, Fabric IQ, Foundry IQ, and Web IQ, further connecting the entire chain of the enterprise AI flywheel and tightly integrating enterprise business data, model orchestration, and application distribution. Morgan Stanley believes that the more models available on the market, the higher the strategic value of Microsoft Corporation as the platform layer connecting models with enterprise business workflows.
Copilot Commercialization: Subscription + Usage-Based Billing Dual Driver; Penetration in Non-Coding Scenarios Still Early
Microsoft Corporation has adopted a hybrid business model of subscriptions plus usage-based billing: subscriptions provide a stable base of revenue, while usage-based billing drives incremental revenue growth in line with customers' actual AI workloads. Management disclosed that some large customers already spend more on the Copilot Cowork product than on all other Microsoft Corporation products combined, but the CFO team emphasized that this case is an extreme sample and does not represent the broader customer base; commercialization of the full product suite has been underway for only a short time, and a reliable penetration pace expectation cannot yet be established.
Currently, Copilot acceptance in coding-related workflows is strong, but consumption for broader knowledge-work office scenarios is still in the cultivation stage; growth potential is considerable, but the pace of adoption remains an important uncertain variable.
Management drew an analogy to Azure's early development: when embracing AI agents, enterprises have very prominent demands for cost controllability and spending observability. Microsoft Corporation launched Agent 365 business analytics and cost management tools that tie AI spending to business outcomes and budget alerting capabilities, helping enterprises manage fluctuating AI usage and clearing obstacles to scaled deployment; however, the direct revenue from these tools depends on the pace at which enterprises expand production-grade AI workloads.
Custom Chip Strategy: Maia and Cobalt as Efficiency Supplements, Not Migrating All Workloads to Custom Hardware
Microsoft Corporation's custom chips are one link in a diversified compute fleet, not an effort to migrate all business to self-developed hardware. Taking the Maia AI chip and the newly released Cobalt 200 as examples, Cobalt 200 targets agentic AI workloads and delivers up to 50% higher performance than the previous-generation Cobalt 100. Microsoft Corporation will continue to purchase chips from external suppliers; the goal of custom chips is to improve performance and cost in specific scenarios as a supplementary lever for overall compute efficiency, rather than an exclusive replacement, continuing the earlier approach of evolving compute chips from a single Intel Corporation (INTC.US) toward diversification across AMD (AMD.US) and Cobalt.
Related Articles

CITIC SEC: Cost reduction and profit growth; the economic inflection point for domestic standalone energy storage is approaching.

FIN STREET SEC (01476): Chairman Zhu Yanhui is under investigation and subject to retention measures

China Securities Co.,Ltd.: AI advanced packaging drives upgrading of glass substrate materials; equipment takes the lead in the early stage of industrialization.
CITIC SEC: Cost reduction and profit growth; the economic inflection point for domestic standalone energy storage is approaching.

FIN STREET SEC (01476): Chairman Zhu Yanhui is under investigation and subject to retention measures

China Securities Co.,Ltd.: AI advanced packaging drives upgrading of glass substrate materials; equipment takes the lead in the early stage of industrialization.






