2026.20: Shifting Alliances in a Changing World

2026-05-15 · Source: Stratechery by Ben Thompson · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

This week's Stratechery brief highlights significant shifts in AI computing, U.S.-China relations, and Amazon's long-term strategic investments. Ben Thompson argues for a new "agentic inference" category in AI, where humans are out of the loop, contrasting with current "answer inference" and impacting architecture trade-offs, potentially favoring China and space over Nvidia. Andrew Sharp discusses the Anthropic-xAI deal, examining its implications for both companies and Elon Musk's ongoing lawsuit with OpenAI. The brief also covers a U.S. presidential visit to Beijing, noting underwhelming deliverables but offering insights into the broader U.S.-China relationship. Amazon's strategy of building infrastructure for its own use then offering it as a service, exemplified by Amazon Supply Chain Services and Project Kuiper (Leo), is presented as a durable model for future growth, particularly in the context of AI's evolving demands.

Key takeaway

For AI Architects and Investors assessing future compute strategies, recognize that the shift to "agentic inference" fundamentally alters hardware requirements, favoring disaggregated resources and custom silicon over tightly coupled GPU clusters. Your investment decisions should prioritize providers like Amazon that have made long-term, capital-intensive bets on internal chip development and power infrastructure, as these will offer sustainable cost advantages and neutrality in a compute-constrained world, reducing dependency on single vendors like Nvidia.

Key insights

Agentic inference, where AI operates autonomously, will redefine compute architectures and market dynamics.

Principles

• Sustainable profits in commodity markets stem from structurally cheaper costs.
• Long-term infrastructure investments yield compounding benefits over time.
• AI's impact necessitates top-down implementation for effective deployment.

Method

Amazon's strategy involves building primitives for internal use, justifying massive expenditure, then selling these services to third parties to increase return to scale and deepen its market moat.

In practice

• Consider agentic inference for future AI architecture planning.
• Evaluate cloud providers based on their custom silicon and networking for inference.
• Invest in power infrastructure to reduce downstream chip costs.

Topics

• Agentic Inference
• AWS Cloud Strategy
• Amazon Custom Chips
• Amazon Supply Chain Services
• US-China Geopolitics

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

Related on AIssential

• not much happened today — The AI market is shifting towards inference economics, agentic platforms, and diverse hardware architectures beyond NVIDIA.
• TAI #197: Anthropic Turned the OpenClaw Demand Signal Into a Product — Agentic AI advancements are rapidly increasing compute demand, straining a constrained global AI supply chain.
• The AI Agent Is Taking Over The Computer — AI agents are merging with computers, transforming interfaces from clicks to goal-driven intelligence.
• Amazon’s Durability — Amazon's strategy involves long-term, capital-intensive investments in infrastructure, initially for internal use, then externalized as services.
• 2026.23: Power Shifts — Elon Musk's companies transform market rules through scale and investor belief, exemplified by SpaceX's space data center vision.
• Agentic AI: We have passed the inflection point — The AI landscape is shifting towards agentic AI amidst a critical data center shortage and rising compute costs.

Open in AIssential →

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