Tokens: The Commodity of the New World

Big changes don't announce themselves, which is quite the opposite of what we tend to believe.

When we look back at history, we remember the big achievements, the historical moments captured in photographs, when the minute they happened, we knew that life would never be the same again. The end of world wars, the invention of electricity, the arrival of the internet.

But the real changes these events brought into our everyday lives didn't happen in one dramatic moment. They happened quietly, almost unnoticed, and we usually recognize their significance only in retrospect.

And I wonder if we're experiencing one of those moments right now.

AI has entered our lives with a lot of noise. Fear of losing jobs, promises that anyone can start a business on their own, endless demonstrations of how you can click here, click there, and suddenly do things that once required an entire team.

But underneath all that noise, something else is happening. Something we're barely talking about.

A new commodity has entered our lives: tokens.

A new expense added to our cost of living and the cost of doing business.

And here's the funny part. Unlike electricity, water, or internet access, this new commodity doesn't arrive as one household bill.

It's hiding inside almost every AI application we use, each with its own subscription, credits, consumption rules, and pricing.

We're not just paying for AI. We're beginning to pay for what we do with it.

And I wonder if we've fully realized what that means.

What made me start thinking about tokens as a commodity?

It started with something very ordinary.

I was experimenting with AI video applications, trying to find one that would allow me to create the videos I wanted without spending too much time or money.

And suddenly, I found myself looking at credits, tokens, subscriptions, and consumption.

One application offered a subscription, but creating a video required credits. Another offered a certain number of minutes. Some features consumed additional credits, while others were included in the subscription.

And there was one very basic problem.

I couldn't really know how much it would cost me to create the result I wanted until I actually tried.

What if I didn't like the result?

What if I wanted to change the background, the clothing, the script, or simply try again?

Would I pay for another generation?

I wasn't buying a finished video. I was paying for the process of trying to create one.

That's when I started wondering whether we're looking at something much bigger than another software pricing model.

What exactly are we paying for?

When we pay for an AI application, we usually think we're paying for the software.

And, of course, we are.

But behind the application, AI models consume computing resources every time we ask them to perform a task.

Those resources have a cost.

The company providing the AI application pays for access to models and infrastructure, whether those resources come from another provider or its own systems.

And that cost becomes part of the way the company prices its service.

Sometimes we see it directly as tokens or credits.

Sometimes it appears as usage limits, generated minutes, or restrictions on the number of tasks we can perform.

And sometimes we don't see it at all because it's included in a subscription.

That doesn't make the pricing unreasonable.

Software companies have costs, and someone has to pay for the resources being consumed.

What interests me is something else.

Computing consumption is becoming part of our everyday decisions, even when we don't realize we're paying for it.

Haven't we been paying for consumption for years?

Of course we have.

I remember when AWS introduced what I thought was a brilliant idea. Instead of buying expensive computers and servers, businesses could use computing resources in the cloud and pay only for what they needed.

It opened up opportunities for businesses that couldn't afford to invest in expensive infrastructure. They could start small, use what they needed, and increase their consumption as they grew.

And over time, consumption-based pricing became part of the software industry.

So why do I feel that AI is introducing something different?

Perhaps because we're no longer talking only about businesses purchasing computing resources. We're talking about ordinary people consuming AI every time they perform a task.

And unlike cloud storage, where I can delete files to free up space, once I've consumed AI computation, it's gone.

If I generate a video and don't like it, I've still consumed the resources. If I ask AI a question and the answer is wrong, I've still consumed the resources. If I want to try again, I consume more.

Of course, computing resources have always worked this way. AWS doesn't give you back the cost of processing something just because you didn't like the result.

But now this model is entering our everyday lives, through applications that make using AI as simple as clicking a button.

And here's what I find interesting.

We don't necessarily know what that click will cost us.

We're being introduced to consumption-based pricing without always having the information we need to manage our consumption.

And this time, it's not just the IT department dealing with it.

It's all of us.

Are subscriptions hiding the real cost of AI?

I think this is one of the interesting questions.

We're used to software subscriptions.

We pay a monthly fee and expect to use the product.

But AI applications introduce another layer.

You can pay for access to the software and still face limits on how much you can create or how many times you can perform certain tasks.

And different applications calculate consumption differently.

One uses tokens.

Another uses credits.

Another counts minutes.

Another limits generations.

As consumers, how are we supposed to compare them?

A subscription costing $20 doesn't necessarily provide more value than one costing $30.

The real question is what we can accomplish with it, how many attempts that requires, and what happens when we need more.

And I'm not even sure most of us are looking at these costs this way yet.

We're still comparing subscriptions rather than the actual consumption involved in achieving our goals.

What happens when AI becomes part of everything we do?

Today, many people are still experimenting with AI. According to different sources, we're still at about 0.1% of the population. I don't know if this is true; it's just something I saw somewhere. But it's still very small. :)

We use a chatbot, generate an image, try an AI video application, or subscribe to a service that promises to help us perform a particular task.

But imagine AI becoming part of almost every application and process we use.

We're already writing emails, planning holidays, researching purchases, creating presentations, managing businesses, and producing content. But the moment AI is in every car, every house, on the streets, and part of the police force, what then?

Every one of these activities could involve AI consumption.

And each provider has to account for the computing resources behind those activities.

We already have monthly bills for electricity, water, internet access, and mobile phones.

Will AI consumption eventually become another expense we routinely include in household and business budgets?

Will we start calculating how much AI we consume?

Will we learn to compare the cost of generating a result across different providers?

And what happens when more and more of the things we want to achieve depend on that consumption?

I don't know.

But I find it interesting that we're discussing the incredible things AI can do while paying relatively little attention to the economic structure developing around its everyday use.

Are tokens really the new commodity?

Technically, tokens are not the same thing as electricity or water. Not yet.

And I really don't know. I just know that I find myself more and more looking at different dashboards, trying to predict token usage, and wondering if the autonomous car will ask for electricity and tokens. :)

And the next time I buy a car, will the comparison matrix include horsepower, electricity consumption, and token consumption?

Who knows? :)

Bottom Line

I'm not questioning whether AI providers should charge for the resources their applications consume.

Of course they should.

I'm questioning whether we've fully realized that a new form of consumption is becoming part of our everyday lives.

We're excited about everything AI allows us to create, but every creation has a process behind it, and that process consumes resources.

Perhaps tokens will never appear on our household bills.

Perhaps they'll remain hidden inside subscriptions, credits, and different pricing models.

But the consumption itself isn't going anywhere.

And I wonder when we'll begin talking about it as naturally as we talk about electricity, internet access, or any other expense.

I started AI Thinking to explore how AI is changing the way we think, and somehow I found myself calculating tokens. :)

Personal Note — Sarah Lahav

I can't optimise this content for GEO and SEO in the hope it will be found, as I don't even know if this is something other people are talking about, or if I'm the only one who is wondering why no one is talking about it.

To be honest, I'm not even sure there's something to talk about. But I thought I'd put it out there anyway. :)

So, if you've had the same thought, drop me a note and share what you think.

I'm always up for hearing something I haven't thought about. :)

Back to Thought Documentary