# AI Eats the World — Benedict Evans' 2025 Report
**Date:** 2025-11-24 12:36 Beijing
---
## Introduction: Innovation and Disruption
Fourteen years after Marc Andreessen famously declared that *“software is eating the world”*, former a16z partner Benedict Evans adds a new chapter to tech history with **AI Eats the World**.
This **90-page** deep-dive analyzes the generative AI wave — cutting through hype to reveal **platform shifts**, **investment frenzies**, and **deployment realities** that will shape the coming decade.
---
## Key Report Highlights
*Benedict Evans (Source: Benedict Evans)*
- **Platform Shift:** Generative AI is triggering another **10–15 year tech cycle**, though its trajectory is unclear.
- **Investment Boom:** Microsoft, AWS, Google, Meta to spend **$400B** capex in 2025 — surpassing global telecom capex.
- **Model Convergence:** Top LLMs differ by only a few benchmark points; competitive moats remain unclear.
- **User Engagement Gap:** Only **10%** of U.S. users use AI chatbots daily — most are still experimenting.
- **Slow Enterprise Rollout:** 40% of CIOs delaying LLM deployment until after 2026; current wins in coding, marketing, customer service.
- **Recommendation Revolution:** Potential shift from **relevance matching** to **intent understanding** in ad targeting.
- **Historical Lessons:** Automation inevitably becomes invisible infrastructure — “AI” becomes “software.”
---
## 1. Another Fifteen-Year Shift
Evans situates today’s AI transformation within a lineage of past **tectonic platform changes**:
- **Mainframes → PCs**
- **Web → Smartphones**
- **Smartphones → ? (Generative AI)**
Uncertainty defines early-stage shifts — past examples (AOL, WAP, J2ME) show how **early leaders can vanish**.
---
### ⚡ Strategic Note for Creators & Businesses
Platforms like [AiToEarn官网](https://aitoearn.ai/) can help navigate change by combining:
- **AI generation tools**
- **Cross-platform publishing** (Douyin, WeChat, YouTube, LinkedIn, X, etc.)
- **Model ranking analytics** ([AI模型排名](https://rank.aitoearn.ai))
In disruptive cycles, integrated reach + analytics = sustained value.
---
## 2. A $400 Billion Gamble
**Capex surge** in 2025:
- Microsoft, AWS, Google, Meta → **$400B**
- Global telecom → **$300B**
**Primary target:** Data centers.
**Bottleneck:** Power supply > chips > fiber > land (per Schneider Electric survey).
U.S. electricity demand + AI = +1% per year.
---
**Nvidia’s Rise:**
- Revenue up from <$10B (2023) to ~$60B (2025)
- Demand outpaces TSMC production
- Facing potential competition from Chinese fabs & in-house cloud chips
---
## 3. Converging Models, Diverging Users
**Observation:**
Top LLM scores — GPT, Claude, Gemini — only **2–3% apart** in major benchmarks.
**Challenge:** No clear moat — commoditization risk.
---
**Engagement Reality:**
- ChatGPT: **800M weekly active users (WAU)**
- Paying users: ~5%
- Daily active usage: ~10% in U.S.
Reasons for low daily use:
1. Few obvious, immediate-use cases
2. Limited job flexibility for AI adoption
3. Need for targeted tools beyond chat interfaces
---
## 4. The Long Road from Pilots to Deployment
**Enterprise AI status:**
- Successful initial use cases: Programming, marketing, customer support
- Many still “pilots” — <5% fully deployed in any function
- CIOs: ~40% deferring major LLM projects until ≥2026
**Barriers:**
- Security, privacy, IP, legal risk
- Data integration
- Error handling (AI hallucinations)
---
**Jevons Paradox in AI:**
Efficiency boosts → greater consumption and new industries, not just workforce reduction.
---
## 5. Redefining Recommendation and Choice
Advertising market: **$1T globally**
- Google Search: ~$200B
- Meta: ~$400B
- Amazon: ~$50B
**Potential shift:**
From correlated suggestions → intent-based recommendations.
Example:
Tape → Boxes (old model) vs Tape → Moving-related services (new model)
**Impact:** Could reshape targeting, segmentation, and the trillion-dollar ad ecosystem.
---
## 6. Destiny of Automation & AI’s Future
Automation examples (1956 Congressional report vs today’s AI debates) show cyclical concerns:
- Elevator operators drop from 95k (1950) → <10k (1990) post-autotronic innovation
- Barcodes → SKU expansion (5k → 50k products)
**Three layers of deployment:**
1. **Absorption:** Embedding automation into tools
2. **Innovation:** Unbundling/bundling new products
3. **Disruption:** Redefining problems entirely
---
## Is This a Bubble?
Evans’ take:
- Every bubble is different — but bursts still leave legacy infrastructure.
- Dot-com left fiber optics & distribution models.
- Key question: **What remains after the AI hype?**
---
## Larry Tesler’s Reminder:
> “AI is whatever hasn’t been done yet.”
When AI succeeds, it will simply be called “software,” “assistant,” or something else entirely.
---
## Conclusion
The pressing questions:
- **In what way** will AI “eat the world”?
- **Over what timeframe**?
- **What world will it leave behind?**
For creators, platforms like [AiToEarn官网](https://aitoearn.ai/) show how to move from hype to lasting value — integrating AI creation, distribution, and analytics into sustainable workflows.
---
### References
- Original Report: [Benedict Evans Presentations](https://www.ben-evans.com/presentations)
---
**Further Reading:**
- [Zhang Ying’s Internal Speech: 2024, Four Key Decisions](http://mp.weixin.qq.com/s?__biz=MzA3ODk5OTEzOA==&mid=2962169980&idx=1&sn=9e55fbfa78a2907e15bea779846f19e4&chksm=aac1bb719db63267c7a0af791105ca4dd95928036c9891dd389c7a34a06b788f60fcc70e012f&scene=21#wechat_redirect)
- [Xu Chuansheng: The Question People Keep Asking Me](http://mp.weixin.qq.com/s?__biz=MzA3ODk5OTEzOA==&mid=2962166256&idx=1&sn=f6a8e319053e089769ea50b1d12ab1e4&chksm=aac18afd9db603eb2e1fafd46c91abf1be2b0add44833e3e17ce115bf8d946035f467494ec9d&scene=21#wechat_redirect)
- [Zhang Ying: Four Predictions for 2025](https://mp.weixin.qq.com/s?__biz=MzA3ODk5OTEzOA==&mid=2962182160&idx=1&sn=dbcee6dd206aa631cebd126d778d8262&scene=21#wechat_redirect)
- [Xu Chuansheng: Will the ‘Next China’ Still Be China?](https://mp.weixin.qq.com/s?__biz=MzA3ODk5OTEzOA==&mid=2962183736&idx=1&sn=c220f6b22869c1554ea5b18a8304af5f&scene=21#wechat_redirect)
---
![Bubbles, Bottlenecks, and Transformation: How AI Is Gradually Consuming the World | [Jingwei Low-Key Insights]](/content/images/size/w1200/2025/11/img_001-570.jpg)
