Is the utopia we expected AI to bring becoming a reality?

Do you remember when ChatGPT was first released to the world in late 2022? Anyone could access it with a simple sign-up, and it was free. Tasks that once required a lawyer to review a contract, a translator to handle foreign documents, or a teacher to explain a concept could now be solved by anyone, alone, at two in the morning. People were ecstatic.

That enthusiasm held something more than mere convenience; it was the sense that AI technology could belong to everyone. Visionaries like Elon Musk went a step further, arguing that AI would cause human productivity to explode, with the resulting surplus distributed evenly across society. They envisioned a world where a child in a poor country could receive world-class education, and anyone with an idea could start a business, regardless of their financial status. Some were even bolder, suggesting that if productivity exploded enough, the very concept of money might one day fade away.

It didn't sound wrong. At times, it even seemed like AI was working exactly that way. But as 2025 passed and 2026 began, something began to change—quietly, yet unmistakably.

At first, it felt like a minor shift. OpenAI limited free users' access to GPT-4 to a few times a day. Google reduced the image generation limit for its free Gemini plan from three images to two. The notices were always similar: "Temporarily limited due to explosive demand," or "A measure to ensure stable service." These were understandable reasons. What could you do if the servers were struggling? I thought it would get better if I just waited, even if it was a little inconvenient.

But the restrictions were never lifted. If anything, they have been steadily tightening, little by little.

Anthropic blocked external automation tool integrations in Claude and lowered peak-time usage limits. xAI’s Grok moved its image generation feature exclusively to paid subscribers. Free features are disappearing one by one, replaced by subscription plans: $20 a month, $200 a month, or custom pricing for enterprises. If you want to use a better model, you have to pay more.

The companies' explanations were partially true. OpenAI is expected to spend about $5 billion in operating costs this year alone. The computational cost of creating a single video is equivalent to processing hundreds of text prompts. GPUs consume electricity, and electricity costs money. AI was never a free resource like air; it wasn't that way from the start. Companies simply hid those costs in the beginning to attract users and capture the market.

That strategy worked, and now the bill is arriving. For us.

Up to this point, it could be read as simple 'business logic.' There’s no such thing as a free lunch, and so on. But let’s consider the near future—a world where AI is as much a part of daily life as our smartphones.

One person uses a free plan. Their daily token allowance is fixed. After clearing a few emails in the morning and drafting meeting materials, they hit their limit before the afternoon. Even if they have more questions, they have to stop. They can start again tomorrow. It’s inconvenient, but there’s no other choice.

Another person has no such constraints. They talk to AI from the moment they wake up. They organize their daily tasks, summarize documents they need to read, and analyze pending decisions from multiple angles. If they hit a wall, they ask immediately. If they see they are heading in the wrong direction, they correct it instantly. The AI runs alongside them at the speed of their own thoughts.

Both live in the same era, the same AI age. But by the end of the day, the difference in what they have produced is significant. That difference compounds daily. A year passes, then five. At some point, the two are standing in completely different places. And both know that the distance isn't due to a difference in individual ability, but a difference in the tokens they can afford to use. And as paid AI models advance, that gap will only widen.

Tokens: the smallest unit of information processing in the AI world. The basic unit of computational resources consumed every time you ask a question and receive an answer.

This word has begun to take on a new meaning. We once spoke of the 'digital divide'—the gap between those with internet access and those without. That gap has largely narrowed thanks to the proliferation of smartphones and the expansion of communication infrastructure. However, the divide in the age of AI is not simply about 'accessibility.' It is a question of how deeply and how freely one can use these tools. Under the illusion that we are all using the same technology, we are, in reality, living in entirely different versions of the future.

Let’s recall that scene from 2022 that everyone was so excited about. Anyone could access it, there was no cost, and it felt like the world was going to change. That feeling wasn't entirely wrong. AI is indeed changing the world. It’s just that the direction of that change might be heading somewhere a little different than what we imagined.

It is not a utopia where surplus is distributed evenly, but a world where those who possess computational resources move faster and further. It is not a world where the concept of money disappears, but one where a new currency called 'tokens' emerges. And the amount of tokens one possesses becomes a power more potent than any previous asset.

We are standing at the threshold of that world right now.

Is the utopia we expected AI to bring becoming a reality?