Exploring the impact of the emerging AI epoch, led by DALL-E, Midjourney, Stable Diffusion, and ChatGPT, on Apple, Amazon, Meta, Google, and Microsoft
The story of 2022 was the emergence of AI, first with image generation models, including DALL-E, MidJourney, and the open source Stable Diffusion …
Context & Ripple Effects
This January 2023 Stratechery piece lands a few months after its own analysis of DALL-E 2 and zero-marginal-cost content generation, which framed image models as a threat to metaverse and social-network economics rather than a curiosity. The new question is bigger: with Stable Diffusion open-sourced and ChatGPT suddenly mainstream, the five platform companies that owned the last computing epoch — Apple, Amazon, Meta, Google, Microsoft — all face the same disruption at once.
The subsequent coverage traces how that played out: AWS moved within months to position itself as a neutral host for rival foundation models, and by mid-2025 the landscape had reorganized enough that the same five companies were being reassessed against OpenAI and Anthropic rather than each other.
First-order effects
- Microsoft and Google, whose search and productivity franchises sit closest to conversational AI, face immediate pressure to ship generative features into their own distribution channels before ChatGPT becomes the interface users reach for first.
- Meta and Apple, whose businesses monetize attention and devices rather than cloud intelligence, are affected differently: zero-cost generated content threatens feed economics on one side, while on-device constraints complicate shipping comparable AI experiences.
Second-order effects
- AWS's decision to offer Anthropic, Stability AI, and AI21 models turns model makers into suppliers competing for platform placement, forcing OpenAI toward exclusive distribution deals — with Microsoft as the obvious anchor tenant.
- If generated content arrives at zero marginal cost, the ad-funded feeds of Meta and Google absorb a supply shock, pushing them toward provenance labeling and human-creator differentiation to protect inventory value.
Third-order effects
- The endgame visible in later coverage is a contest over defaults rather than models: by 2026 the race for the 'default AI' has usage splintering across products, meaning whoever owns entry points — OS, browser, cloud console — captures the value regardless of whose model wins.
- Sustaining that competition requires compute at a scale where capacity itself is strategy: Epoch AI's projection that AI chips in use will double every nine months makes capital access, not research talent, the binding constraint separating the five incumbents from everyone else.
The trend: Generative AI is shifting competition from who builds the best model to who controls distribution and compute, reordering the Big Tech pecking order around platform position rather than research leadership.