100 Trillion Examples (And One Secret)

One trillion of anything is a lot. But we can’t wrap our brain around large numbers like this. So we have to find analogies to help us to grasp the scale. Relationships to time or money or space can help. An example that’s stuck with me expresses the scale of one trillion in terms of something we can all relate to – one second in time and one dollar.

We each understand intuitively what spending one dollar feels like. We understand what 10 seconds feels like; what holding our breath for 30 seconds feels like. But spending a dollar and a large number of seconds isn’t intuitive until we express them in days or years. So if you spent one dollar every second:

• One million dollars would take 12 days to spend.

• One billion dollars would take 31 years to spend.

• One trillion dollars would take 31,700 years to spend.

Keep this in mind the next time you read about Apple or Nvidia having a market capitalization of $4 trillion dollars. Or when combined Canadian government debt exceeds $2.3 trillion.

Now, consider the scale of the research that OpenRouter released last week. They analyzed 100 trillion tokens (a token is a word or a snippet of a word used by LLMs) or about two weeks worth of real-world AI usage. That’s the equivalent of reading every one of the 40 million books in the U.S. Library of Congress 25 times.

This isn't a little survey of 500 tech bros on Twitter!

Why is this important? This 100 trillion token analysis isn’t abstract data. It’s like that old saying: “Don’t tell me about your values, show me your budget.” How people and companies are spending their “token budget” reveals their real priorities. They’re investing time and money into AI tools in ways that show where and how they’re getting real value. This is a map of how humans are “spending” these tokens on a global scale.

But when you look at that map, you see something interesting. At first glance, it looks like most of the world is just playing games. The data shows that a massive percentage - over 50% in some models - is categorized as “Roleplay” or recreational chat.

This creates a “Cloud of Noise.” Which makes it easy for the media and pundits to dismiss AI as a toy for hobbyists. But that cloud of “fun stuff” is masking the real story.

Hidden behind that wall of “spending” on recreational chat, the data shows a massive and intentional shift in professional usage. Categories like “Programming,” “Reasoning,” and “Agentic Work” are growing fast. People aren't just playing. They’re also doing real work. And that usage is growing fast (programming has grown from 11% to over 50%).

Furthermore, the share of total tokens routed through reasoning-optimized models has climbed sharply, moving from a negligible slice in early 2025 to now exceeding fifty percent of all usage. Prompt and output length also increased dramatically – from under 2,000 tokens to more than 5,000. Prompts are getting more detailed and complex. And people are getting better, and more useful, answers.

This data snapshot tell us that the “Hype Phase” is ending, and the “Utility Phase” is taking over. The data shows a shift from simple “chatting” to complex “reasoning” and “coding.” People aren't just playing with these tools anymore. They are building and growing businesses with them.

For an SMB or nonprofit, this means that while the headlines focus on fun and games, the real opportunity is sitting in the background. While most are still chatting recreationally, you can start leveraging these tools today to solve real, everyday problems.

But there’s a catch. While the usage data is screaming “Growth,” some other human sourced data is whispering “Fear.”

A separate report from Anthropic this week revealed the results of interviews with more than 1,200 professionals. But they did something clever: they didn't use human interviewers. They used an AI to conduct the interviews.

Why? For speed and efficiency and because humans often shade the truth to other humans to save face. We’re often more candid with machines because machines don't judge us.

The result of these honest conversations? 86% of respondents think AI is saving them time and helping them to be more productive. But 69% of professionals admitted they are afraid of “negative judgment” if colleagues find out they use AI.

This creates an interesting paradox:

• The Hard Data (the 100 trillion tokens) show that usage is exploding.

• The Soft Data (the interviews) shows that most users don’t want to talk about it.

As an Alberta SMB or nonprofit leader, you might feel the same fear. But here’s the opportunity: most people are already (and often covertly) using these tools. I've written about this before - we're still in the early innings of this cognitive revolution. You're not alone but if you're reading this, you're already ahead of many businesses.

The "Secret Centaur" Opportunity

If you are an SMB owner or nonprofit leader in Alberta, this is your “Arbitrage Moment.” Most of your competitors are either:

• Not using AI because they don't get it.

• Using it secretly, but afraid to operationalize it because of the stigma.

This means you can take a simple, low-profile step forward while others hesitate.

The “Glass Slipper” Effect

The 100 Trillion token study found one other important insight for SMBs and nonprofit leaders: users churn through models until they find the one specific tool that fits their workflow - a phenomenon they called the “Glass Slipper” effect. Once they find that fit, they stop looking and start building.

My advice this week:

Ignore the stigma. The data proves you aren't alone - you are just part of the silent majority. Don't try to master all of “AI.” Just look for your Glass Slipper. Find one “boring” problem - one pain point - scheduling, marketing, research, invoicing, drafting emails - that a tool like Gemini, Claude, NotebookLM or ChatGPT can help solve. Work with it. Build with it. Experiment. Once you find it and the shoe fits, you’ll never look back.

Jeff Uhlich

CEO & Founder, augmentus inc.

December 7th, 2025