Everyone is waiting for Nvidia, but the truly underestimated one might be Amazon

Over the past year, Generative AI has not only been the hottest topic in the tech world but has almost defined the main narrative of global capital markets. In this sweeping wave of intelligence, Nvidia has become the king among "water sellers" thanks to its monopoly on GPU computing power, Microsoft has embedded Copilot into every corner of office scenarios by partnering with OpenAI, and Google, leveraging its deep technical foundation, is attempting to reshape search and its ecosystem. These three companies have become the absolute protagonists under the spotlight, with their stock prices repeatedly reaching new highs and their valuation logic completely reconstructed.

However, amidst this lively AI feast, another giant, equally deeply embedded in the AI infrastructure, data, applications, and commercialization cycle, has remained unusually "quiet" for quite some time—that is Amazon. Despite possessing the world's largest cloud computing platform, AWS, and the largest e-commerce retail network, Amazon seems to have become a forgotten corner in the AI trading logic of the secondary market. Its stock performance, while steady, has not enjoyed the valuation premium experienced by Nvidia or Microsoft.

Against this backdrop, Morgan Stanley recently released a somewhat "counter-consensus" research report. Maintaining an "overweight" rating on Amazon with a price target of $300, the report clearly points out that AWS and the retail business are being systematically underestimated by the market as beneficiaries of Generative AI. This judgment, like a stone thrown into a calm lake, has sparked deep thought in the investment community.

This raises the question: in today's highly crowded AI narrative, is Amazon's "quietness" a reflection of market rationality, or is it a missed opportunity? Is today's Amazon a rare "buy-the-dip" window, or a mature giant, repackaged by the AI narrative but lacking explosive potential? To answer this, we need to cut through the market noise and re-examine Amazon's true ecological position in the AI era.

Judging solely by the sentiment and narrative logic of the secondary market, Amazon does not really come across as a typical core AI asset. In investors' stereotypes, AI investment often seeks "purity" and "explosiveness." Nvidia has an incredibly clear "computing power monopoly" logic, with every wave of large model iteration directly translating into demand for H100, B200, and other chips; Microsoft, on the other hand, offers the alluring product story of "OpenAI + Copilot," giving the market a vision of exponential growth in software subscription revenue.

By contrast, Amazon's AI story appears fragmented, slow, and even somewhat clunky. It has not released a consumer-facing model like Sora or GPT-4 that has sparked nationwide discussion; its AI strategy is more hidden in B-side services and back-end efficiency optimization. This "seemingly unsexy" characteristic is the starting point for Amazon being undervalued.

For a long time, the market has priced Amazon based on two old narratives. The first is the "retail empire": despite its huge revenue, the e-commerce business has low profit margins and requires heavy logistics capital expenditure, being seen as a "grunt work." The second is the "cloud computing concerns": although AWS is the industry leader, growth has slowed in recent quarters and it faces fierce competition from Microsoft Azure and Google GCP, with fears its cloud market share in the AI era will be eroded. Within this old framework, AI is seen more as "the icing on the cake" for Amazon—used to optimize logistics or assist in code writing—rather than as a core variable for structural revaluation.

However, this perception overlooks a key fact: Generative AI is not an industry that wins solely on model capability, but one that is highly dependent on computing power, data, distribution scenarios, and a commercial closed loop. Large model training requires massive computing power, inference needs stable low-latency infrastructure, application deployment needs rich data scenarios, and commercialization requires mature payment and user systems.

In these four dimensions, Amazon precisely possesses scale advantages, and these advantages are latent. Unlike companies driven by a single blockbuster model or hardware product, Amazon's AI capability is more like "water, electricity, and gas." It does not directly sell "intelligence"; it sells the "soil" that generates intelligence. The market underestimates Amazon because it is accustomed to looking for "gold rushers" and ignores the "infrastructure provider" supplying shovels, maps, tents, and transport ships by the river. This lag in perception is exactly the root cause of the expectation gap.

What truly led Morgan Stanley to turn bullish on Amazon again is not whether it has the "strongest model," but its position in the AI commercialization chain. As AI moves from "technical demonstrations" to "industrial implementation," Amazon's AWS and retail business form the most solid "monetization machine" in the AI era.

First, there's Amazon Web Services (AWS). In the era of Generative AI, the role of cloud computing has fundamentally shifted from traditional "IT outsourcing" to being a "computing power factory + model hosting platform." Every step of large model training, inference, fine-tuning, and deployment consumes real computing resources. For the vast majority of enterprises, building their own computing clusters is too costly and technologically challenging, so moving to the cloud is the only choice. This is exactly where AWS excels at monetization.

Unlike Microsoft Azure, which leans toward "software-bound AI" (strongly promoting its own models and Office integration), or Google, which emphasizes "technology first" (highlighting model capability), AWS's greatest strength lies in its "neutrality" and "economies of scale." In the AI era, enterprise customers are increasingly concerned about data privacy and vendor lock-in. Through the Bedrock platform, AWS allows customers to freely choose third-party models from Anthropic, Meta, AI21, etc., while also supporting enterprises to build their own models. Whether it's startups or traditional giants, they ultimately need stable, low-latency, and scalable computing power and cloud services. This makes AWS the beneficiary of the AI wave that is "least dependent on the success or failure of any single model." No matter which large model wins out, as long as AI traffic grows, AWS's bills will grow too. Additionally, Amazon's in-house Trainium and Inferentia chips are providing customers with more cost-effective computing options, further consolidating its cost moat.

What is more easily overlooked is the relationship between Amazon's retail business and AI. The market often narrowly interprets AI's application in retail as "customer service chatbots," but this is just the tip of the iceberg. Generative AI's ROI (return on investment) in supply chain forecasting, inventory optimization, advertising placement, and personalized recommendations is much higher than that of "demonstration AI."

Amazon possesses one of the world's most complex and data-dense retail systems. From user clicks, browsing, adding to cart, payment, logistics, to after-sales, every link generates massive amounts of data. Each efficiency improvement brought by AI can almost directly translate into improved profit margins here. For example, using AI for more accurate regional sales forecasts can reduce inventory backlogs and allocation costs; generative AI can optimize product detail pages and advertising materials to increase conversion rates; AI can optimize logistics routes to reduce last-mile delivery costs.

While the market is still debating "when will AI make money" or "who is the next killer app," Amazon has quietly figured out this equation internally. The retail business's operating leverage is extremely high; once AI-driven efficiency gains cover fixed costs, the resulting free cash flow will be astonishing. This kind of "silent profit release" may not be as exciting as launching a new model, but is more deterministic and sustainable in financial models.

Given Amazon's deep AI foundation, is now the right time to enter? The real issue is not whether Amazon "has AI," but whether, as AI moves from narrative to realization, the market will reprice this deterministic cash flow model.

From a valuation perspective, Amazon is not currently enjoying a premium commensurate with its AI infrastructure status. Compared to AI hardware companies that are highly dependent on single product cycles or capital expenditure, Amazon's risk is lower, but its imagination is also more heavily suppressed. The market gives Nvidia a "growth stock" valuation, Microsoft a "monopoly software" valuation, and Amazon still suffers from a "traditional retail" discount.

That is the source of the divergence:

Bears believe that Amazon is too large, and the incremental revenue from AI can only bring "marginal improvement" in the face of trillions in revenue, making it hard to drive a substantial stock price increase. They worry about bottlenecks in cloud business growth and the fragility of the retail business amid macroeconomic fluctuations.

Bulls—including Morgan Stanley—are betting that once AWS and retail both enter the AI-driven profit release stage, the market will be forced to reprice Amazon with a new model. This is not only an increase in revenue, but also a structural expansion of profit margins. The rising proportion of high-margin AI services in AWS, combined with cost reductions brought by AI on the retail side, will create the "Davis Double Whammy."

Therefore, the core of "whether you can buy the dip" does not depend on short-term stock price volatility, nor on whether the next quarterly report beats expectations, but on a fundamental investment philosophy judgment:

Do you believe the biggest winners in AI are the "dream sellers" or the "bill collectors"?

"Dream sellers" rely on technological breakthroughs and vision, with greater stock price volatility, higher upside but also higher risk; "bill collectors" rely on infrastructure and ecosystem dominance—regardless of which upper-layer application wins or loses, they take a cut. In the early stages of AI development, the market chases dreams; but in the mid-stage of large-scale AI deployment, cash flow and certainty will become scarce assets.

If it's the latter, then today’s Amazon may be standing at a neglected starting point. It does not need to top every model benchmark; it only needs to ensure that every AI application running on its cloud pays it, and that every transaction achieved through its recommendation system contributes profit. This business model has not become obsolete in the AI era—on the contrary, the explosive demand for computing power and data due to AI makes it even more solid.

Looking back at the history of tech stocks, every technological revolution undergoes a process of "tech explosion - bubble formation - separating the real from the fake - value return." Currently, we are in the stage of transitioning from bubble formation to separating the real from the fake. The market is beginning to realize that not every company labeled with AI will survive, but those that control computing power, data, and scenarios will see their value rise as AI penetration increases.

What makes Amazon unique is that it is both a king of the old era (e-commerce and logistics) and a cornerstone of the new era (cloud and AI). This dual identity makes it seem ambiguous in the capital markets, but it is this very ambiguity that provides a margin of safety. Morgan Stanley's $300 price target is not just a number but a confirmation of Amazon's transition from "growth anxiety" to "profit release" logic.

For investors, focusing on Amazon is no longer about FOMO (fear of missing out) but rather a rational calculation of value distribution in the AI industry chain. When the tide goes out, the true giants collecting the bills will ultimately prove their worth. Amazon may not double in the short term like Nvidia, but what it offers is a more certain compound return in the long-distance AI race. As AI consensus stocks reach their peak, looking back at this silent giant may well mark the beginning of a return to rationality in capital markets.