Why AI Is so Centralized: How It's Built, Who Controls It, and What Comes Next

Why It Matters

Decentralizing AI with crypto‑based trust could democratize access, curb monopolistic power, and accelerate innovation across industries.

Key Takeaways

• AI infrastructure is dominated by a few centralized companies.
• Decentralized compute can improve efficiency by leveraging everyday devices.
• Crypto provides programmatic trust for secure machine-to-machine interactions.
• Autonomous AI agents will become sovereign economic actors within a year.
• Shifting AI to open infrastructure could reshape power dynamics and innovation.

Summary

The video examines the growing centralization of artificial intelligence, arguing that today’s AI services are controlled by a handful of dominant firms and that this concentration poses both technical and philosophical risks. It proposes a shift toward decentralized infrastructure that leverages idle compute on everyday devices, promising greater resource efficiency and a more equitable distribution of power.

The hosts explain that decentralization requires solving verification challenges absent in centralized systems, and that crypto offers a programmatic trust layer essential for machine‑to‑machine contracts. They trace the evolution of model training—from supervised labeling to unsupervised and reinforcement learning—showing how reduced human input accelerates autonomy.

Key remarks include the forecast that “machines will be autonomous, sovereign economic actors…within the next 12 months,” and the analogy that AI, like early social media, is becoming fundamental human infrastructure. The discussion underscores that current centralization mirrors the early internet’s power concentration.

If AI moves to open, decentralized protocols, control over models, data, and compute could shift away from tech giants, fostering a Darwinian market for intelligence, lowering costs, and democratizing innovation. This transition would reshape regulatory, economic, and societal dynamics, making AI’s future less dependent on a few corporate gatekeepers.

Original Description

A few big companies control most of the infrastructure behind AI.

Most people experience AI through a wide range of different apps that actually depend on a deeply centralized stack of data and compute. In this conversation, Ben Fielding and Harry Grieve — cofounders of decentralized machine learning protocol Gensyn — explain why this matters, and what it would take to rebuild AI as open infrastructure instead.

From unused global compute to the philosophical implications of machine intelligence, they argue that the next evolution of AI must be owned, coordinated, and verified in a fundamentally different way.

Highlights

00:00 – Intro

00:29 – The biggest misconception about AI infrastructure

01:20 – Why centralization in AI is a deeper problem than people realize

04:19 – Why AI needs crypto

05:51 – How AI models are trained

08:15 – The rise of autonomous AI agents with onchain identities

10:37 – Lightning round

Ben Fielding on X: https://x.com/benfielding

Harry Grieve on X: https://x.com/harrygrieve

Gensyn on X: https://x.com/gensynai

Follow a16z crypto on X: https://x.com/a16zcrypto

Subscribe for more news and updates: a16zcrypto.substack.com