This is your last chance. After this, there is no turning back. You take the red pill, you stay in Wonderland and I show you how deep the rabbit hole goes.

Morpheus, The Matrix
Based on a letter by Ribbit Capital

THE TOKEN
REVOLUTION

Tokens are making the world legible to machines, encoding who we are, what we own, what we do, and what we want. Every business on the planet will be defined by this over the next decade.

500T+
Tokens / Month (Google, I/O 2025)
Up from ~10T at I/O 2024, a 50x jump in a single year
1,000x
Inference Cost Collapse
The cost to run frontier AI dropped 1,000x in roughly 3 years
150x
Payment Tokens Issued (Visa)
From 100M tokens in 2014 to 16B+ today, payments are token-by-default
Framework from Ribbit Capital's June 2025 Token Letter. Stat values updated to current.
Tokenization = Rendering the World Legible for ComputersEvery Business Is Becoming a Token FactoryInference Costs Collapsed 1,000x in 3 YearsProprietary Data Is the New MoatStablecoins Are the First Machine-Native MoneyKYA (Know Your Agent) Will Replace KYCVertical Token Systems Will Dismantle BPOs and ERPsUnassuming Data Capturers Will Be AI's First Killer AppsAgents Thrive When Given Identity, Data, and ResourcesThe Race to Own Memory Tokens Is Wide OpenTokenization = Rendering the World Legible for ComputersEvery Business Is Becoming a Token FactoryInference Costs Collapsed 1,000x in 3 YearsProprietary Data Is the New MoatStablecoins Are the First Machine-Native MoneyKYA (Know Your Agent) Will Replace KYCVertical Token Systems Will Dismantle BPOs and ERPsUnassuming Data Capturers Will Be AI's First Killer AppsAgents Thrive When Given Identity, Data, and ResourcesThe Race to Own Memory Tokens Is Wide Open
๐Ÿ”ค
The Basics

Not All Tokens Are
Created Equal

Tokenization is defined as rendering the world legible for computers, encoding who we are, what we own, what we do, and what we want, so that software can act on it. Click each token type to understand what it is, why it matters, and who is winning the race to own it.

Tokenization is the process of rendering the world for computers.Ribbit Capital Token Letter, June 2025
โšก
Why Now

6 Forces That Made
This Moment Inevitable

Six structural forces converged to make the Token Revolution inevitable. These are not passing trends. Each one is self-reinforcing and continues to compound.

๐Ÿ“‰
Running Cost: Collapsing
1,000x
Inference cost dropped 1,000x in three years. Open-source models like Llama 4 and DeepSeek V3.1 broke the hyperscaler monopoly. Any founder can now run complex AI tasks without breaking the bank.
๐Ÿ› ๏ธ
Build Cost: Plummeting
100x
LangChain, AutoGen, CrewAI compress months of engineering into hours. AI coding agents now author the majority of new code at frontier companies. Vibe coding is the default.
๐Ÿ’พ
Raw Inputs: Exploding
23% CAGR
Global data sphere growing from 106 ZB (2022) toward 291 ZB (2027). AI transcription costs $0.05โ€“0.20/min vs $1โ€“3 for humans, a 5โ€“60x collapse. Every signal is becoming capturable.
๐Ÿ”—
Infrastructure: Opening
<$10
Launching a token on Solana or Base still costs less than $10. Transaction fees below $0.02. Stripe, Visa, NPCI now ship agent-ready payment rails as default.
๐Ÿš€
Builders: Arriving
1M+
Over a million developers shipping with agent frameworks. 3/4 of recent YC batches are AI-native. The fastest-growing repos on GitHub are LLM orchestration, MCP servers, and agent toolchains.
๐Ÿ
The Race: Everywhere
1,500T+
Google + Microsoft alone now process 1,500T+ tokens/month combined, up from 500T just a year ago. Every layer of the stack โ€” compute, models, apps โ€” is in an all-out sprint. Trillions in equity value have already shifted.
The Scale of What Has Already Happened
Refreshed for mid-2026 ยท framework from the June 2025 letter
$300B
Stablecoin AUM
$3T/mo
On-chain settlement volume
$15B
Tether annual profit
900M
ChatGPT weekly users
$800B
Lost from poor customer service (US)
1.4B
Aadhaar identity tokens (India)
$400B
Annual BPO market (to be disrupted)
$800B+
Annual fintech technology spend
๐Ÿญ
The Core Concept

Every Business Is Becoming
a Token Factory

Token factories transform tokens into more valuable tokens, or into higher levels of intelligence. Every software business can be modeled as supplying, building, or orchestrating token factories.

Suppliers
Data &
Identity
Companies holding proprietary tokens, biometrics, financial history, behavioral data. Think Plaid (782M linked accounts), Stripe (billions of payment tokens), ID.me (140M users), CLEAR (30M users). These moats took years to build and can't be replicated even by teams armed with coding agents.
Builders
Models &
Agents
Token factories that transform raw tokens into intelligence. From foundation models (OpenAI, Anthropic, Google) to specialized factories built on proprietary data, Abridge for healthcare, Harvey for law, AlphaSense for finance. The litmus test: given the same inputs, can your factory produce more valuable outputs?
Orchestrators
Platforms &
Agents
Agents source, orchestrate, and direct tokens to deliver outcomes for users, like 3PL integrators. Vertical Token Systems will replace ERPs and BPOs: starting with simple wedges, growing to become the nervous systems of entire industries. SafetyKit, Decagon, and Harvey are early examples.
The biggest factor in evaluating businesses for the coming decade may be the tokens they stockpile, create, or collect every day.Ribbit Capital Token Letter, June 2025
๐Ÿ“Š
The Data

The Cost Collapsed,
The Volume Exploded

Two trajectories pointing at the same future: AI is becoming radically cheaper while machine-native transaction volume runs away from human volume.

Inference Cost Per Million Tokens ($)
AI Tokens Processed (Trillions / month)
๐Ÿ’ก
Strong Beliefs, Loosely Held

10 Bets on
the Token Future

Three acts: sourcing tokens to feed the machines, building agents that feed on them, and supplying the fuel that powers it all. Click any belief to expand.

Sourcing TokensFeeding the Machines
Building AgentsHow Machines Feed Us
Supplying FuelWhat Powers the Machines
๐Ÿ”„
The Moat

Token Moats Are Built
on Trust Flywheels

Data is becoming more like money, the more you have, the better positioned you are to create more. The most valuable companies of the next decade will compound by harnessing these self-reinforcing loops.

๐Ÿ’ฐ
Value Loop
Asset Tokens
More Liquidity & Acceptance
More Trust & Brand
More Usage
Bitcoin ยท Tether ยท Visa
๐Ÿง 
Expertise Loop
Expert + Knowledge Tokens
More Inference
Better Predictions + More Trust
More Distribution
Harvey ยท Abridge ยท AlphaSense
๐Ÿ‘ค
Personalization Loop
Identity / Context / Memory
More Engagement
Richer Data + More Trust
Better Personalization
ChatGPT Memory ยท Plaid ยท ID.me
๐ŸŽฏ
Test Yourself

What's Your Token
Revolution IQ?

Think you absorbed all of it? Prove it.

TOKEN REVOLUTION IQ TEST
Question 1 of 6
How much has the cost of AI inference dropped over the last three years?
๐Ÿงญ
Investment Thesis

Investment Themes
from the Letter

Ten areas the letter calls out as the most important bets in the token economy over the next few years.

01
Teams treating agents as a first-class customer, becoming default parts of the "agent stack" (e.g. standard MCP server for payments or identity)
02
Enterprise Vertical Token Systems in compliance, tax, accounting, wealth, healthcare, legal, insurance, and real estate
03
Teams imagining financial services for agents, how they'll pay, earn, borrow, and own assets in an agentic economy
04
Reinventing customer relationship management for an agent-first world, across every touch point, not just support
05
Identity solutions for a world of agents, giving people control and ownership of their data, especially memory tokens
06
Teams working toward the universal personal finance assistant, "George Bailey for everyone, agents that know you, never sleep, and don't want a Porsche"
07
Giving superpowers to creators and small business owners, 300M+ SMBs globally need more customers and access to enterprise-grade tools
08
Tokenized asset infrastructure, issuers, processors, and admins. Bridge analogs will emerge across every asset class and geography
09
"Newspapers for machines", prediction markets, oracle networks, tokenized content that trusted agents can use as reliable information sources
10
Resources for autonomous agents: compute/data markets, governance and reputation systems, real-world token factories with defensible physical IP