AMZN — Earnings Review Q4 FY25 (Phil Fisher)

Date: 2026-04-01 Quarter: Q4 FY25 (Dec-2025) | Reported: Feb 5, 2026 Market cap: $2,200B | P/E (GAAP): 28x | Revenue growth: 13.6% YoY FY25 Revenue: $716.9B | FY25 EPS: $7.17 | TTM FCF: $11.2B

Verdict

I have spent a career observing companies at inflection points — moments when management must choose between the comfortable path and the bold one. Amazon under Andy Jassy in the fourth quarter of fiscal 2025 presents one of the clearest cases I have encountered of a company that is "fortunate because it is able." The distinction matters enormously.

AWS re-accelerated to 24% year-over-year growth — the fastest in thirteen quarters — on a $142 billion annualized revenue base. Advertising grew 23% to $21.3 billion quarterly. North America retail achieved a record 9.0% operating margin. These are not the results of a company riding an industry tailwind; they are the results of a management team that has methodically built infrastructure advantages — custom silicon, regionalized fulfillment, an AI platform stack — that competitors cannot replicate at comparable economics.

The $200 billion capital expenditure plan for fiscal 2026 will strike many observers as reckless. I believe it is the opposite. It is the kind of long-range investment that sacrifices current free cash flow — compressed to $11.2 billion from $38.2 billion — for a position that will generate returns for a decade or more. The $244 billion AWS backlog, growing 40% year-over-year, tells us the demand is real. The question is not whether Amazon should be spending; it is whether they can execute the buildout without eroding the operational excellence that made AWS dominant.

My scuttlebutt research, however, reveals a concern that the financial statements cannot surface: AWS is experiencing significant talent attrition among its most experienced engineers, with 69-81% regretted departure rates reported. The October 2025 DynamoDB outage, which took an abnormally long time to diagnose, may be an early symptom. For a company whose competitive moat rests fundamentally on operational excellence, this is the single most important thing to monitor.

Assessment: Outstanding company entering a transformative investment cycle. Hold through the capex compression. The growth arc is intact and widening. But watch the people — Fisher's Point 7 (labor relations) and Point 9 (management depth) bear close scrutiny.

Management Quality Assessment

Transparency Under Pressure (Point 14)

Mr. Jassy was remarkably forthcoming on the earnings call about the magnitude of the investment commitment. He stated plainly: "We expect to invest about $200 billion in capital expenditures across Amazon, but predominantly in AWS." He explained the rationale clearly — demand exceeds supply, capacity is monetized as fast as it is installed, and the AI opportunity transforms every customer experience.

Where management was less forthcoming — and this I note carefully — was on the question of financial guardrails. Two experienced analysts (Mark Mahaney of Evercore and Doug Anmuth of JPMorgan) pressed directly for a free cash flow floor or minimum return threshold on the capex program. Both times, management deflected. Mr. Olsavsky spoke of "strong return on invested capital" without quantifying it. Mr. Jassy pivoted to the demand narrative. This is not quite "clamming up," but it is a notable absence of specificity from a management team that is otherwise impressively detailed.

I have found over many years that when management is unwilling to commit to financial boundaries, it usually means one of two things: either they genuinely believe the opportunity is so large that setting limits would be irrational, or they are uncertain about the returns. Given the $244 billion backlog and the 24% growth rate on $142 billion, I am inclined toward the first interpretation. But I will be watching for this to change.

Promises vs. Delivery

Management has delivered consistently on every operational commitment made in prior quarters. This pattern of promise-keeping is one of the strongest indicators of management quality that I track.

Long-Range Thinking (Point 12)

The $200 billion capex commitment is perhaps the most consequential example of long-range thinking in modern business history. Management is deliberately compressing free cash flow from $38.2 billion to 11.2billion—andguidingtowardfurthercompression—becausetheybelievetheAIinfrastructureopportunityjustifiesmulti − yearinvestment.TheyaresimultaneouslyinvestinginKuipersatellites( 1 billion incremental quarterly cost), quick commerce expansion internationally, and 100+ new Whole Foods stores.

This is textbook Fisher. The companies that produce the most spectacular long-term returns are precisely those whose management will sacrifice current profits to build competitive advantages that compound for years. The key question is always whether the sacrifice is genuine investment or mere expenditure. The evidence here — custom chips delivering 30-40% better price-performance, capacity monetized immediately upon installation, backlog growing faster than revenue — points strongly toward genuine investment.

R&D and Growth Pipeline (Points 1-3)

Is R&D Coordinated and Effective?

Amazon's research and development efforts are perhaps the most coordinated I have observed in a company of this size. The custom silicon program illustrates this beautifully:

• Trainium2: 1.4 million chips landed — the fastest ramping chip in Amazon's history. Powers the majority of Bedrock inference. Fully subscribed. Multibillion-dollar annualized run rate.
• Trainium3: Already in production workloads. Nearly all supply committed by mid-2026.
• Trainium4: Expected 2027 launch with 6x FP4 compute improvement over Trainium3.

The cadence is remarkable — each generation is developed while the prior generation is still scaling. This is not a "crash program" (which I have always warned against). It is systematic, disciplined development coordinated across research, manufacturing, and sales. The results per dollar are outstanding: the custom chips business achieved >$10 billion ARR with triple-digit growth, delivering 30-40% better price-performance than comparable GPUs. My scuttlebutt research confirmed that SemiAnalysis identified structural supply chain advantages — 23% component discounts via Astera Labs warrants, rapid CoWoS-to-rack deployment, and a cableless design that reduces assembly failures.

New Growth Avenues

The hallmark of an outstanding company (Point 2) is that management creates new growth avenues as existing ones mature. Amazon currently has multiple growth vectors at different stages of development:

This depth of pipeline is exceptional. When the current AI infrastructure buildout matures, the revenue from agents, Bedrock, and the satellite business will likely be providing the next growth wave. I have rarely seen a company with this many genuine growth opportunities developing simultaneously.

Fifteen Points Assessment

Summary: 13 of 15 points rate strong to excellent. Point 7 (labor relations) is the sole genuine concern, and it warrants close monitoring. Point 14 (transparency) has a minor blemish on the FCF guardrail question. No disqualifiers.

Scuttlebutt Findings

My scuttlebutt investigation — sourced from competitors, customers, employees, industry analysts, and product benchmarks — reveals the following mosaic:

What Competitors Say

AWS maintains its #1 position in cloud infrastructure at approximately 31-33% market share (Synergy Research, Q4 2025). Azure is at 24-26%, GCP at 11-12%. AWS's 24% growth rate on the largest base means the absolute dollar gap is widening, not narrowing. This is the mark of a company that is "fortunate because it is able" — its dominance is not an accident of timing but a function of relentless execution.

However, a notable signal from AWS re:Invent 2025: Amazon made three strategic reversals — accepting multi-cloud (AWS Interconnect), on-premises AI infrastructure (AI Factories), and democratized model training (Nova Forge at $100K/year). These are pragmatic responses to competitive pressure. They signal that AWS management is not rigidly ideological, which I consider a positive sign of adaptability.

What Customers Say

Enterprise customers continue to migrate to AWS. The customer win list announced this quarter — OpenAI, Visa, NBA, BlackRock, U.S. Air Force, and CrowdStrike — represents a breadth across industries that suggests the competitive moat is widening, not narrowing. More top 500 US startups use AWS as their primary cloud provider than the next two combined.

A concerning data point: only 4.3% of Y Combinator's 2024 cohort uses Bedrock, versus 88% using OpenAI directly. This suggests that while AWS wins on infrastructure, it is losing the developer hearts-and-minds war for AI application layers. This pattern — startups adopting competitors' AI tools then building on their infrastructure — creates a risk that workloads could migrate over 3-5 years. However, inference economics (where Trainium excels) may keep the infrastructure moat intact even if the application layer is contested.

What Employees Say

This is where my concern concentrates. Internal documents indicate 69-81% regretted attrition among key AWS engineers. The October 2025 DynamoDB outage — which took an abnormally long time to diagnose — may be symptomatic of institutional knowledge loss. Glassdoor reviews show strong compensation but persistent concerns about work-life balance and the pace of organizational change. The return-to-office mandates in 2025 caused meaningful attrition.

I have written at length about the importance of labor relations (Point 7). A company that loses its most experienced engineers is losing something that cannot be replaced by capital expenditure. The $200 billion in capex builds data centers and installs chips — but the operational excellence that differentiates AWS from Azure and GCP lives in the institutional knowledge of the people who built those systems. This warrants close monitoring in coming quarters.

Product & Technology Quality

Trainium2 delivers 30-40% better price-performance than comparable GPUs for inference workloads, validated by Anthropic training Claude on the platform (Project Rainier, 500K+ chips). However, AWS compute pricing ($43.80/mo) trails Azure OpenAI on inference latency (Bedrock 245ms vs Azure 180ms time-to-first-token). The latency gap matters for real-time applications but is likely addressable through Trainium3 improvements.

The Prime membership moat is formidable: 220 million global members, 2% churn rate, estimated annual value of $1,430 versus $139 cost — a 10.3x value-to-cost ratio. JPMorgan projects 350 million members and estimates a $20 price increase would not cause significant churn. One of the stickiest consumer franchises in modern commerce.

Management Reputation

Andy Jassy is four years into the CEO role and has clearly established his strategic vision: the cost optimization period (2022-2023), the profitability transformation (operating margin from 1.8% to 11.7%), and now the AI infrastructure pivot. The $200 billion capex guidance is the boldest bet of his tenure. His track record of delivering on operational commitments lends credibility to the forward plan.

(Sources: Synergy Research Group, SemiAnalysis, Glassdoor, Y Combinator cohort data, JPMorgan Prime analysis, AWS re:Invent 2025 reports, industry benchmarks)

Four Dimensions of Conservative Investing

Dimension 1: Operational Excellence — Outstanding

Amazon achieves the rare feat of simultaneously improving delivery speed and reducing cost-to-serve for three consecutive years. The regionalized fulfillment network, 1 million+ robots, same-day delivery reaching 100 million US customers, and 70% increase in same-day items — all while lowering unit costs — represents operational excellence at a scale I have rarely encountered.

Dimension 2: The People Factor — Strong with a Blemish

The management team is deep and capable. Jassy has delegated effectively across retail, AWS, advertising, devices, and new ventures. Internal promotions are the norm. The blemish is the AWS talent attrition issue. If Point 7 deteriorates further, this dimension would require downgrade.

Dimension 3: Business Characteristics — Dominant

What can Amazon do that others cannot do about as well? Build custom silicon optimized for their own cloud workloads. Operate a logistics network that delivers same-day to 100 million customers at declining cost. Run the world's largest cloud platform, the third-largest advertising platform, and a 220-million-member subscription ecosystem — all reinforcing one another through data and distribution advantages. The flywheel is the moat.

Dimension 4: Price — Reasonable

At 28x trailing GAAP earnings on a business generating 30% EPS growth, the PEG ratio is 0.93 — below the threshold I would consider overvalued. The P/Operating Cash Flow of 15.8x is genuinely modest for this quality of business. The sum-of-parts analysis (AWS alone at $2.1 trillion at cloud-comparable multiples) suggests the market is assigning substantial discount for the capex cycle and regulatory overhang. When the capex cycle normalizes — likely 2027-2028 — and operating cash flow of $140 billion+ translates to free cash flow, the current market capitalization will look conservative.

Growth Arc Analysis

Amazon is emphatically "fortunate because it is able." The evidence:

• AWS did not merely benefit from the cloud migration trend — Amazon created the cloud computing market and continues to define its frontiers through custom silicon, AI platforms, and satellite connectivity.
• The advertising business was built by leveraging proprietary shopping data that no competitor possesses — it was not an obvious adjacency, but management saw the opportunity and built it into a $69 billion business.
• Same-day delivery was achieved not through spending more on logistics but by re-engineering the entire fulfillment network around regional hubs — an operational innovation that competitors cannot replicate at comparable cost.

The multi-year growth runway remains extensive. Cloud computing is estimated at 10-15% of enterprise workload penetration. AI adds an entirely new S-curve on top of the migration S-curve. US e-commerce is 22% of total retail. Advertising is shifting from linear TV to streaming. Kuiper opens satellite connectivity as a new market. Each of these runways extends for years.

Revenue growth has re-accelerated from 8.7% (Q1 FY25) to 13.6% (Q4 FY25), driven by AWS's inflection from 17% to 24%. The sequential revenue adds are widening quarter by quarter — $0.5 billion to $2.6 billion in AWS alone over the last four quarters. The $244 billion AWS backlog growing at 40% versus 24% revenue growth creates a widening bullish divergence that strongly suggests sustained 20%+ AWS growth through fiscal 2026.

What I Am Watching

1. AWS talent retention — The 69-81% regretted attrition rate is the most important leading indicator. If it does not improve within two quarters, the operational excellence moat is at risk. Look for management commentary on retention initiatives and whether service reliability metrics (uptime, incident resolution time) deteriorate.

2. AWS operating margin trajectory — Down from 39.5% (Q1 FY25) to 35.0% (Q4 FY25). As $200 billion in capex flows through depreciation, margins could compress further into the low-30s. The key is whether Trainium's cost advantage offsets the depreciation burden. A sustained move below 30% would be a concern.

3. Bedrock competitive position — The 4.3% YC adoption versus 88% for OpenAI is a startup adoption gap. If Bedrock does not gain share in the developer ecosystem, the AI application layer could become a structural advantage for competitors, even if AWS wins on infrastructure.

4. Capex-to-revenue ratio — FY2026 capex at ~$200 billion would be approximately 28% of estimated revenue. This is unprecedented. If demand signals weaken — watch for backlog growth deceleration or customer pushback on pricing — the return on this investment comes into question.

5. Agentic AI impact on advertising — Two analysts flagged the risk that AI agents bypassing the Amazon search bar could compress the advertising funnel. Jassy pivoted to Rufus but did not quantify the risk. This deserves monitoring as agent adoption scales.

Conclusion

Amazon in Q4 FY25 is the rare company that satisfies nearly all fifteen of my criteria for an outstanding investment. The management team has demonstrated remarkable competence across an extraordinarily complex business. The growth runway is among the longest I have observed — multiple secular tailwinds, each with years of penetration ahead. The competitive moat is widening through proprietary silicon, logistics infrastructure, and the Prime ecosystem. The valuation at 28x earnings with a PEG below 1.0 is reasonable for this quality.

I agree broadly with Atlas's assessment but add significant nuance on the people factor. Atlas rates conviction at 4/5 on capex execution risk. I rate it similarly, but my primary concern is different: it is not the capital that worries me, but the people deploying it. The AWS talent attrition data is the kind of signal that financial statements cannot surface but scuttlebutt reveals — and it is precisely the kind of early warning that separates thorough analysis from superficial number-reading.

If the job has been correctly done — if the investor has identified a company that is "fortunate because it is able," with outstanding management, multiple growth avenues, and a widening competitive moat — the time to sell is almost never. Amazon qualifies. The $200 billion capex cycle will compress free cash flow for one to two more years, and there will be quarters when the market loses patience. Those quarters will be opportunities, not reasons to sell.

Hold with conviction. Monitor Point 7 (talent retention) closely.

Analysis by Philip A. Fisher. First analysis for AMZN. No position disclosure.