Why is Microsoft currently the best target for "bottom-fishing" in AI? Goldman Sachs: AI profit margins will recreate the cloud era's expansion miracle.

Microsoft's stock price has fallen 16% from its peak in mid-December last year, and it is still struggling to recover the key $500 threshold. For traders on the sidelines, the core question is simple: does this correction present an exceptional “bottom-fishing” opportunity? Goldman Sachs analyst Gabriela Borges, after visiting Microsoft’s Redmond headquarters and having in-depth conversations with executives, gave a highly affirmative answer. Goldman Sachs’ conclusion is straightforward: **Among all the tech giants under its coverage, Microsoft is the “best candidate” for achieving compounded growth through the AI product cycle.** This judgement is not just based on a vague vision, but is backed by a clear financial trajectory—Goldman expects Microsoft’s earnings per share (EPS) to steadily approach $35 by fiscal year 2030, representing a compound growth rate of over 20%. For investors, this means that the current panic may well be an entry ticket, as Microsoft is establishing its dominance in the age of AI through the flexibility of its infrastructure and unique profit-margin advantages. ## **Repeating the Cloud Era Miracle: The Inevitable Path from High Cost to High Profit** The core logic presented by Microsoft’s management to Goldman Sachs is that **the current AI cycle bears an astonishing similarity to the early cycle of Cloud.** Investors need not fear the currently high cost of AI investment. Looking back to the early cloud transformation stage, it was likewise flooded with high costs and weak unit economics, but as scale effects, utilization rates, and engineering efficiency improved, profit margins expanded significantly. **During the cloud transformation, Microsoft set mid-term gross margin targets and kept actual results within 100 basis points of its target every year—this precise execution is set to repeat in the AI era.** **In fact, Microsoft believes its leadership in the AI cycle even surpasses that of the cloud era.** This confidence stems from an almost fanatical discipline and operational efficiency. For example, Microsoft discovered inefficient elements in a model that caused excess compute usage; in the cloud era, solving such an issue might have taken 2-3 months, but in today’s urgency around AI operations, the team delivered an optimization within a single weekend. As scale grows, Microsoft is confident that the core cloud business margins will further expand, and AI business margins will improve over time. ## **The Essence of the Moat: Gross Margin Advantage from OpenAI Partnership and LLM Abstraction Layer** **On the software level, Microsoft’s competitive advantage is transforming into tangible financial barriers.** Goldman Sachs pinpoints that **Microsoft’s partnership with OpenAI gives it a unique gross margin advantage.** Because Microsoft holds intellectual property (IP) rights over OpenAI models, it does not have to pay additional API fees when calling these models, in effect waiving a huge “gross margin tax” compared to other software providers—a significant competitive edge. Furthermore, Microsoft is redefining the role of large language models (LLMs). Just as virtual machines abstracted hardware and containers abstracted operating systems, Microsoft views LLMs as the next-generation abstraction layer, able to abstract the logic of applications themselves. Future applications will not rely on hard-coded rules, but will shift toward intent-driven execution. Microsoft’s Foundry platform has the opportunity to become the central control layer for routing, governance, and cost optimization. While the market currently focuses on the rising absolute costs of new models (e.g., ChatGPT 5.2 vs. 5.1), Microsoft points out that next-generation models are being designed to be more efficient. In the future, as token costs decline, value will increasingly concentrate at the platform layer, and LLM-related costs of goods sold (COGS) will become trivial. ## **The “Generality” Strategy in Infrastructure: Refusing Customer-Supplied Chips to Maintain Control** **In infrastructure construction, Microsoft demonstrates strong strategic resolve, refusing seemingly tempting short-term compromises.** Management made it clear that “Bring Your Own Chip” (BYOC) models offer neither economic appeal nor strategic advantage for Microsoft. BYOC isolates the infrastructure stack and undermines the driving force of cloud margins: scaled procurement, full-stack integration, and end-to-end optimization. **Microsoft’s profit advantage comes from holistic optimization across data centers, power, cooling, networking, and silicon layers—not any single component.** Thus, Microsoft did not reach a BYOC agreement with Anthropic, but insisted on leveraging its own purchasing and balance sheet advantages to provide clients with required chip architectures, preserving overall system efficiency. The core of this strategy is “generality.” Microsoft adopts a “delayed binding” strategy in data center design and supply chain, postponing design and deployment decisions as long as possible to retain flexibility. For example, its new “Fairwater” design employs a two-level structure and 3D rack layout, shortening cable length to boost GPU performance. To achieve agility in switching workloads and chip types, Microsoft is even willing to sacrifice small performance gains that would be achieved through customized cooling or chip design. This strategy enables Microsoft to flexibly allocate capacity between training and inference tasks based on demand signals, minimizing utilization risk. ## **Enterprise Adoption Inflection: Shift from “Whether to Use” to “When to Scale Up”** On the demand side, Goldman Sachs observes a clear shift in the market’s wind direction. Compared with a year ago, enterprise clients’ discussions around Copilot have shifted from debating ROI and “whether” to adopt, to focusing on “when” and “how much” to adopt. Budget uncertainty is fading, and clients are no longer holding back budgets to hedge macro risks as they did in Q4 last year. **Microsoft notes that enterprise AI adoption is already widespread and shows a “take root, then expand” pattern.** Clients typically start with pilot projects of a few hundred licenses, and as familiarity grows, quickly scale into the thousands. In pricing, Microsoft has adopted a value-based strategy, currently offering a low-priced commercial SKU at $21/user to widen the funnel, but with a long-term goal of, through feature expansion, supporting pricing above $30/user. While many customers are currently experimenting with building AI agents in-house (DIY), Microsoft believes that as the complexity of maintaining models, managing updates, and building reliable connectors compounds over time, clients will eventually return to Microsoft’s platform solutions. **Sales incentives have also shifted, moving from focusing on pricing to accelerating the time for clients to “realize value,” indicating Microsoft’s move from pure sales to deep ecosystem lock-in.** Risk warning and disclaimer The market carries risks; investments should be made with caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions herein suit their particular circumstances. Invest at your own risk.