Amazon’s Durability

2026-05-05 · Source: Stratechery by Ben Thompson · Field: Business & Management — Corporate Strategy & Leadership, Operations & Process Management, E-commerce & Digital Commerce · Depth: Advanced, long

Summary

Amazon.com Inc. recently launched Amazon Supply Chain Services (ASCS), consolidating its existing air and ocean freight, trucking, and last-mile delivery offerings into a new suite for businesses. This move, which saw shares of rivals like FedEx and UPS decline, fulfills a decade-old prediction about Amazon's strategy of building "primitives" for internal use and then offering them to third parties, mirroring the AWS model. The article also details how Amazon's long-term, capital-intensive approach, including its custom Nitro and Graviton chips, positions it favorably in the evolving AI landscape, particularly for inference workloads. While initially seen as ill-prepared for AI training due to networking and GPU supply issues, AWS's disaggregated architecture and investment in its own Trainium chips are now advantageous for the growing inference market, especially for reasoning and agentic AI.

Key takeaway

For CTOs and AI Architects evaluating cloud strategies, Amazon's long-term infrastructure investments and custom silicon strategy for AI inference present a compelling case. Your organization could benefit from AWS's cost advantages and neutrality, especially as AI shifts from training to inference and agentic workloads. Consider AWS for its robust, cost-effective inference capabilities and its ability to support frontier models like Anthropic, potentially reducing dependency on single-vendor GPU solutions.

Key insights

Amazon's strategy involves long-term, capital-intensive investments in infrastructure, initially for internal use, then externalized as services.

Principles

• First-best customer justifies massive infrastructure investment.
• Commodity markets reward structural cost advantages.
• Long-term investments compound benefits over time.

Method

Build infrastructure primitives for internal use, scale them, then offer them as services to external customers to gain leverage on capital costs and deepen market moats.

In practice

• Consider custom silicon for long-term cost advantages.
• Prioritize power infrastructure for AI factories.
• Invest in disaggregated compute for inference workloads.

Topics

• Amazon Supply Chain Services
• AWS Cloud Strategy
• AI Inference Optimization
• Custom AI Chips
• Logistics Network

Best for: CTO, AI Architect, MLOps Engineer, Director of AI/ML, VP of Engineering/Data, Investor

Related on AIssential

• What are Amazon CEO Andy Jassy's Six 'Key Truths' of AI? — AI adoption is accelerating at an unprecedented rate, driving massive investment and competitive shifts.
• Amazon Is Playing Both Sides of the AI Game, and That’s Fine — Tech giants are strategically investing in rival AI firms and developing model-routing services to ensure platform relevance and integrate proprietary offerings.
• Amazon's Big Spending Plans and Bitcoin's Rebound — Massive AI infrastructure investments are reshaping market valuations and driving new business models across tech and media.
• Peter DeSantis: Architect of Amazon’s AI and Silicon Future — Amazon's AI strategy is driven by in-house custom silicon and infrastructure to reduce costs and accelerate innovation.
• The State of The Inference Economy — AI's economic value is shifting from training to inference, which now drives revenue and dominates compute costs.
• Amazon’s Lower Cost AI Chip Offerings Hold Appeal for Businesses — Amazon's lower-cost AI chips aim to attract businesses.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Stratechery by Ben Thompson.