If you've used ChatGPT, Gemini, or Claude recently, you've probably noticed the word 'token' showing up everywhere—in usage dashboards, API billing pages, and model spec sheets. But for most of the 2010s, that same word meant something completely different to millions of people: a digital asset on a blockchain, bought and sold on exchanges alongside altcoins and meme stocks.

A Word With Deep Roots

The term 'token' traces back to Old English tācen (mark, sign), cognate with German Zeichen. Its Proto-Indo-European root *deik- means 'to show or point,' which explains why the word has always meant 'not the thing itself, but what points elsewhere.' This etymological logic made it perfect for both cryptocurrency and AI—two industries that independently latched onto the same word for concepts involving representation and value transfer. In crypto, token evolved specifically because Ethereum's ERC-20 standard (formalized November 2015) needed a term distinguishing on-chain assets without their own blockchain from native currencies like bitcoin. Vitalik Buterin had participated in the 'colored coins' project before founding Ethereum—the naming was almost predetermined. By 2017-2018, the ICO frenzy raised over $13 billion using ERC-20 tokens, and 'token' became virtually synonymous with crypto itself.

The AI Revolution Overwrites the Dictionary

But here's where it gets interesting for anyone watching linguistic shifts in tech. OpenAI's ChatGPT launched in November 2022 and hit 100 million users within two months—the fastest technology adoption in recent memory. While cryptocurrency took twelve years to reach DeFi Summer, AI went from breakthrough to ubiquity in just over three years. The numbers are stark. According to OpenRouter platform data, weekly token call volume across the top ten models surged from approximately 1.24 trillion in March 2025 to roughly 13.95 trillion by February 2026—a tenfold increase in under a year. Goldman Sachs projects a further 24-fold jump in global monthly token consumption between 2026 and 2030, driven by AI agent adoption.

Why the Shift Stuck

The critical difference isn't just speed—it's cultural baggage. Crypto's string of scandals (FTX, Terra/Luna, Three Arrows Capital) attached negative connotations to 'token' in public consciousness. Meanwhile, AI positioned token as something more mundane and useful: a unit of compute you consume when drafting emails, debugging code, or generating reports. When someone outside the crypto world hears 'token' today, their mental frame almost certainly activates an API billing dashboard rather than a price chart. Linguists call this a shift in collocates—the company words keep. A few years ago, token appeared alongside blockchain, wallet, holder, and sale. Today its neighbors are input, output, context window, and usage. The definition remained unchanged ('smallest processing unit'), but what it processes is unrecognizable from a decade ago when tokens were footnotes in compiler manuals.

Key Takeaways

  • 'Token' now activates AI compute associations by default for most people under 35 who weren't already deep in crypto
  • OpenRouter data shows token call volume grew 10x between March 2025 and February 2026, hitting nearly 14 trillion weekly
  • Goldman Sachs projects AI agents will drive a 24-fold increase in token consumption through 2030
  • The word's economic nature (tradeable asset vs. consumable cost) makes both meanings overlap completely in commercial language

The Bottom Line

This is what linguistic displacement looks like in real-time—two massive industries choosing the same ancient English word, one eventually overwriting the other in everyday consciousness. Crypto had a twelve-year head start and still couldn't lock down 'token.' AI arrived later, moved faster, and claimed the default meaning. That's not an accident; it's a lesson in which technology actually lands with regular people.