Amazon’s recent pullback has reignited a familiar debate: is Amazon falling behind in the race to monetize generative AI — or is the company quietly laying the groundwork to reclaim momentum? Investor jitters about AWS’s cloud revenue growth and Microsoft’s head start with OpenAI have pressured the stock, but a prominent Wall Street analyst argues there’s a bullish indicator hidden in Amazon’s capital spending that could change the narrative. This piece untangles the headlines, verifies the key numbers, and analyzes what Amazon’s AI strategy — and its rivalry with Microsoft Azure — means for the company’s next chapter.
The market’s concern centers on two linked ideas: Microsoft has lockstepped into the generative-AI zeitgeist through a deep partnership with OpenAI, and Azure’s rapid AI monetization is outpacing AWS’s visible progress. Microsoft’s integration of OpenAI models into enterprise offerings such as Copilot and Azure OpenAI Service has created a conspicuous revenue path tied directly to AI services. That clarity has helped drive sentiment and price performance for Microsoft in recent quarters.
Meanwhile, AWS’s AI story has been more distributed. Amazon has rolled out its own foundation models and AI services — notably Amazon Bedrock, Titan models, CodeWhisperer, and custom silicon such as Trainium and Inferentia — and it has struck investments and partnerships (Anthropic being the most prominent). But those pieces have not yet produced the same headline-grabbing revenue attribution that Microsoft displays, so the market has reacted to perceived slower cloud revenue growth at AWS compared with Microsoft and Google Cloud.
Investor anxiety intensified after Amazon’s recent quarter: results beat core earnings and revenue, but AWS growth and forward guidance landed below the most bullish expectations. That opened the door for narratives that Amazon’s massive AI spending might not yield the near-term growth investors expect.
Amazon’s defense is that foundational infrastructure builds take longer to show up in revenue, but when they do, they may underwrite a very large addressable market. The Morgan Stanley view treats capex as a leading indicator: large-scale, multi-year investments should precede accelerated AWS growth, particularly as more enterprises move AI from pilot to production.
This creates a bifurcated risk/return profile:
However, the capex path is not a free pass. It raises the bar on execution: AWS must translate raw capacity into differentiated, developer-friendly offerings and secure developer migration paths to Trainium or otherwise neutralize CUDA lock-in friction. It must also demonstrate consistent enterprise adoption of Bedrock and related managed services at scale.
For traders and long-term holders, the pragmatic stance is to treat Amazon as a strategic, long-duration investment in AI and cloud infrastructure: one that requires patience while watching the specific adoption metrics listed above. The most important thing for investors is not to assume capex alone guarantees success, but also not to dismiss capex as mere spending — when paired with demonstrable utilization and product-level monetization, it becomes the kind of durable moat that defines the next phase of cloud competition.
The next 12–24 months will be decisive. If AWS can translate rising square footage and server capacity into predictable, high-utilization AI workloads with developer-friendly tooling, the capex will look prescient. If adoption takes longer or customers consolidate around Microsoft’s packaged AI offerings, Amazon will need to show additional strategic differentiation.
For now, the “bullish indicator” is real: capital commitment at scale. But it’s only a starting gun — the race will be won by execution, developer adoption, and the ability to convert infrastructure into recurring, high-margin cloud AI revenue.
Source: Investor's Business Daily https://www.investors.com/news/technology/amazon-stock-ai-aws-microsoft-azure-gen-ai/
The market’s concern centers on two linked ideas: Microsoft has lockstepped into the generative-AI zeitgeist through a deep partnership with OpenAI, and Azure’s rapid AI monetization is outpacing AWS’s visible progress. Microsoft’s integration of OpenAI models into enterprise offerings such as Copilot and Azure OpenAI Service has created a conspicuous revenue path tied directly to AI services. That clarity has helped drive sentiment and price performance for Microsoft in recent quarters.Meanwhile, AWS’s AI story has been more distributed. Amazon has rolled out its own foundation models and AI services — notably Amazon Bedrock, Titan models, CodeWhisperer, and custom silicon such as Trainium and Inferentia — and it has struck investments and partnerships (Anthropic being the most prominent). But those pieces have not yet produced the same headline-grabbing revenue attribution that Microsoft displays, so the market has reacted to perceived slower cloud revenue growth at AWS compared with Microsoft and Google Cloud.
Investor anxiety intensified after Amazon’s recent quarter: results beat core earnings and revenue, but AWS growth and forward guidance landed below the most bullish expectations. That opened the door for narratives that Amazon’s massive AI spending might not yield the near-term growth investors expect.
Overview: the “bullish indicator” and the capex war
Morgan Stanley analyst Brian Nowak (and other bank analysts) flagged a relatively simple but consequential datapoint: Amazon’s capital expenditure trajectory. The argument runs like this:- Hyperscalers are racing to build AI-ready infrastructure — specialized data centers, server clusters, and optimized networking to host training and inference at scale.
- Amazon has been accelerating capex in 2025, with a large share earmarked for AWS data centers and AI compute.
- That capex boost, when combined with Morgan Stanley’s forward-looking analysis of AWS data center square footage coming online, implies a realistic path for AWS revenue to re-accelerate and regain share of AI-driven cloud spend.
What the numbers say (verified and cross-checked)
Technical claims and numbers matter here. Several key figures drive the debate, and they withstand verification across multiple reporting outlets and filings.Capital expenditures
- Amazon has signaled a very large capex program for 2025. Company commentary and multiple industry reports put Amazon’s 2025 capital outlays well into the dozens of billions — in some reports north of $100 billion for the year across property, equipment, data centers, logistics, and other investments. Company statements and analyst models show a very heavy skew of that spending into AWS infrastructure and AI-capable assets.
- Microsoft’s fiscal 2025 capex plan for AI infrastructure sits in the roughly $80 billion range in public reporting. That number has been widely reported by outlets focusing on hyperscaler spending plans and reflects Microsoft’s own disclosures.
Data center capacity (square footage)
- Analysts estimating AWS’s forward-looking data center footprint used third-party datasets and filings to model expected additions. One prominent analysis projects roughly 8.5 million square feet of incremental AWS data center space to come online in 2025 and about 10 million square feet in 2026, based on project filings, campus proposals, and lease activity.
- AWS’s own filings show dramatic year-on-year increases in operating square footage, consistent with aggressive capacity expansion.
Revenue growth comparisons
- AWS’s recent reported cloud revenue growth rates (mid-teens) have trailed the double-digit/accelerating growth reported by Microsoft Azure and Google Cloud in some recent periods. Public earnings and company revenue disclosures show Microsoft and Google often reporting higher year-over-year growth percentages in the cloud segment during the generative-AI acceleration.
- Analysts point out that AWS’s trailing growth doesn’t negate long-term opportunity; rather, it creates a visible gap investors are fixated on until cloud monetization of AI becomes clearer for AWS.
What Amazon is actually building for AI
Amazon’s strategy is multi-pronged and infrastructure-first. The major elements include:- Amazon Bedrock: a managed service that exposes multiple foundation models (including third-party models) through a single API. Bedrock is designed to let customers pick and customize models while keeping data inside their AWS environment.
- Titan and Nova foundation models: Amazon’s own family of models (Titan and Nova/Nova-like offerings) that are meant to be available for customization and production use on Bedrock.
- Custom silicon: AWS’s Trainium (and Inferentia for inference) chips are intended to offer a lower-cost alternative to GPU-based training/inference for certain workloads. Amazon has promoted Trainium’s cost advantages and in-house performance on specific model types.
- Strategic partnerships and investments: Amazon’s multibillion-dollar involvement with Anthropic is a clear play to secure access and workload demand for AWS. Anthropic has made AWS a primary cloud partner, tying model training and inference workloads to Amazon’s ecosystem.
- Developer and enterprise tooling: CodeWhisperer for code suggestions, QuickSight generative BI features, and integrations into existing AWS data services are all aimed at making it easier to move enterprise data and workflows into gen-AI applications on AWS.
Strengths in Amazon’s playbook
Amazon has several durable advantages that support the bullish capex-to-growth thesis:- Scale and balance-sheet muscle. Few companies can fund sustained, multiyear capex at the scale Amazon projects. That gives AWS the ability to underwrite capacity growth without needing immediate payback.
- End-to-end stack control. Amazon builds and operates its own chips, data centers, and cloud services. That vertical stack reduces dependence on external hardware suppliers and can reduce cost-per-inference/training over time.
- Breadth of model choice. Bedrock’s multi-model approach can attract enterprises that prefer vendor neutrality or want to test multiple foundation models without being locked into a single provider’s model.
- Enterprise relationships and workloads. AWS remains a primary cloud provider for many Fortune-scale enterprises. Its extensive installed base gives Amazon an addressable audience for AI migration once enterprises standardize on production-ready AI services.
- Data center pipeline. Project filings and public plans show an aggressive buildout in key regions (e.g., North Carolina, Virginia), which supports the capacity story underpinning analyst models.
Real risks and execution challenges
The bullish scenario isn’t guaranteed. Several credible risks could slow or limit AWS’s AI monetization:- Timing and utilization risk. Building physical capacity is necessary but not sufficient. Customer migration to production-scale AI workloads — and the timing of that shift — is uncertain. Underutilized infrastructure is a capital drag.
- Supply-chain and power constraints. Chip availability, rack and board supply, and regional power limitations can delay capacity ramp or reduce yield per square foot.
- Software and developer lock-in hurdles. A significant fraction of machine learning workloads today are CUDA-optimized for NVIDIA GPUs. Convincing customers to port models or optimize for Trainium requires tooling, libraries, and a path that doesn’t impose excessive retooling costs. Developer friction could slow adoption of AWS’s custom chips.
- Competitive product bundling. Microsoft’s deep integration of OpenAI and its embedding of generative AI into productivity tools gives it unique routes to revenue that are already proving sticky with customers. Google’s Gemini and end-to-end AI tooling also provide competitive pressure.
- Margin and cost pressures. Massive capex increases can weigh on free cash flow and margins in the short term. Investors demanding near-term returns may penalize perceived overinvestment if revenue acceleration lags.
- Reputational and regulatory headwinds. AI safety, accuracy, and privacy concerns can influence enterprise purchasing behavior and complicate deployments in regulated industries.
Tactical differences: Amazon’s approach vs Microsoft’s approach
A useful way to think about the competition is by contrasting the go-to-market and product strategies:- Microsoft: deep, exclusive partnership with OpenAI; strong enterprise software integration (Copilot across productivity suites); clear revenue paths tied to subscription and per-seat models; aggressive capex and Azure-first cloud positioning for OpenAI workloads.
- Amazon: multi-model neutrality through Bedrock; custom silicon and hardware-first cost advantages; a “platform for many models” strategy that emphasizes choice; large investments in physical capacity and chip development to drive lower unit economics.
The investor calculus: short-term pain vs long-term payoff
For investors, the crux is timing. The market rewards visible, repeatable revenue today; it discounts optionality far into the future. Microsoft’s clear AI revenue streams have a direct, near-term impact on investor sentiment.Amazon’s defense is that foundational infrastructure builds take longer to show up in revenue, but when they do, they may underwrite a very large addressable market. The Morgan Stanley view treats capex as a leading indicator: large-scale, multi-year investments should precede accelerated AWS growth, particularly as more enterprises move AI from pilot to production.
This creates a bifurcated risk/return profile:
- If adoption accelerates on the cadence Nowak and others anticipate, AWS could re-accelerate to mid-to-high-teens or even 20%+ growth, supporting a bullish price target and validating the capex call.
- If demand lags or if Microsoft/Google capture a disproportionate share of enterprise AI spend because of product or go-to-market advantages, Amazon could face a longer, more painful transition with pressure on margins and returns.
What to watch next (actionable signals)
Investors and enterprise customers should track several lead indicators to separate hype from reality:- Deployment and utilization metrics: announcements of major enterprise AI production deployments (not pilots) that explicitly cite AWS/BEDROCK as the platform.
- Trainium adoption signals: tooling, library support, and public migration case studies that show large models running efficiently on Trainium hardware.
- AWS gross margins on cloud infrastructure: improvement would indicate better pricing or higher utilization.
- Incremental revenue per square foot: any company or analyst disclosures that reconcile added capacity to realized revenue would validate the Morgan Stanley framework.
- Strategic partnerships and exclusivity deals: new investments or long-term contracts (like Anthropic) that tether large model providers to AWS will be significant tailwinds.
Verdict: cautious optimism with a clear playbook
Amazon’s stock volatility reflects a classic market tension between narrative and execution. Microsoft’s early monetization wins and the visible OpenAI integration created a compelling short-term narrative, while Amazon’s approach has been quieter and more infrastructural. The Morgan Stanley argument — that capex is the bullish indicator — is credible because hyperscaler economics are fundamentally capacity-driven in the early AI era.However, the capex path is not a free pass. It raises the bar on execution: AWS must translate raw capacity into differentiated, developer-friendly offerings and secure developer migration paths to Trainium or otherwise neutralize CUDA lock-in friction. It must also demonstrate consistent enterprise adoption of Bedrock and related managed services at scale.
For traders and long-term holders, the pragmatic stance is to treat Amazon as a strategic, long-duration investment in AI and cloud infrastructure: one that requires patience while watching the specific adoption metrics listed above. The most important thing for investors is not to assume capex alone guarantees success, but also not to dismiss capex as mere spending — when paired with demonstrable utilization and product-level monetization, it becomes the kind of durable moat that defines the next phase of cloud competition.
Short checklist for tech leaders and IT decision-makers
- If you’re evaluating cloud providers for AI workloads:
- Assess portability costs between CUDA-based stacks and AWS’s Trainium-optimized stack.
- Evaluate Bedrock for model selection flexibility and compliance controls for regulated data.
- Benchmark total cost of ownership (TCO) for training and inference across providers, including expected discounts for long-term commitments.
- If you’re tracking vendor viability:
- Watch for production-level case studies that specify model sizes, performance metrics, and cost differentials.
- Monitor public disclosures about data center utilization and capacity ramp timelines.
Final thoughts
The debate over whether Amazon is “behind” in AI misunderstands the nature of infrastructure-driven competition. Microsoft’s integration and product-led monetization are visible and compelling today. Amazon’s strategy is less flashy but — based on verifiable capex plans, data center build-outs, custom silicon, and major partnerships — plausibly positioned to win a large share of the long-term AI compute market.The next 12–24 months will be decisive. If AWS can translate rising square footage and server capacity into predictable, high-utilization AI workloads with developer-friendly tooling, the capex will look prescient. If adoption takes longer or customers consolidate around Microsoft’s packaged AI offerings, Amazon will need to show additional strategic differentiation.
For now, the “bullish indicator” is real: capital commitment at scale. But it’s only a starting gun — the race will be won by execution, developer adoption, and the ability to convert infrastructure into recurring, high-margin cloud AI revenue.
Source: Investor's Business Daily https://www.investors.com/news/technology/amazon-stock-ai-aws-microsoft-azure-gen-ai/