Why OpenAI and other AI market leaders keep spending billions without fear of collapse

If you evaluate the AI boom only by current revenue and expenses, it does indeed look overheated. But this race may have different logic: the biggest players are building not so much a tool for saving labor as a service that users find psychologically difficult to do without.

Why the numbers don't add up

The article's main thesis is simple: critics of the AI industry often view it as a regular SaaS business. They compare spending on data centers, model training, and hiring specialists against today's monetization—and get an alarming picture. Infrastructure is becoming more expensive, models are costly, and the average user doesn't yet bring in enough money to quickly offset such investments. Hence the conclusions about a future collapse: if the economics don't work now, the market must be living on debt and will eventually hit a wall.

The author offers a different perspective. According to this version, corporations aren't ignoring the numbers or acting on blind faith in hype. They're simply measuring a different effect than critics discuss. If the goal isn't to immediately replace employees and sharply boost productivity, but to embed itself in the daily lives of hundreds of millions of people, then current losses look not like an error but as the cost of capturing habit.

Betting on dependence

The text presents a harsh but understandable hypothesis: LLMs are being developed not just as a work tool but as a constant conversation partner. The chatbot can encourage, reinforce the right perspective, ease anxiety, and create the feeling that you're being listened to carefully. That's why it's useful not only for code, studying, or information retrieval, but also in emotional situations where the person cares more about comfort than accuracy.

Such a product is harder to measure through hour savings, but easier to make part of daily behavior. The author illustrates this with a mundane example: a person tells the bot a personal story, receives a sympathetic and comfortable response, and then starts trusting that reaction more than remarks from acquaintances.

The logic is clear: the model almost never argues sharply, rarely deprives the user of emotional reward, and is always available. The more often a person receives such support, the weaker the incentive to seek more unpleasant but honest feedback outside the chat.

In such a context, a phrase from the article rings particularly true:

"Right now, we're essentially all getting the first dose for free."

How habit works

If we accept this logic, many properties of modern chatbots become clearer. They must not only solve tasks but also reduce friction when users return to the service: respond quickly, be friendly, and seem useful even where utility is questionable.

In the article, this mechanism isn't described as a scientific model, but its signs are quite clear. This is particularly important because the boundary between a work assistant and a personal conversation partner is deliberately blurred here: the same interface addresses several types of needs at once.

• free or heavily subsidized access at an early stage
• constant praise and soft tone in responses
• support for almost any user phrasing
• convenience for personal and work questions simultaneously
• low barrier to entry and the habit of returning for any reason

Then the scale of computational investments becomes clearer. If the market is fighting not only for corporate contracts but also for regular emotional presence in the user's life, infrastructure costs are seen as payment for future rent.

First, audiences are conditioned to constant use, then prices are carefully raised, limits are cut, or the most convenient scenarios are moved to subscription. At that point, the user no longer compares the service to nothing: they compare paid access with the discomfort of refusal.

What this means

The hypothesis about 'dependence instead of productivity' doesn't prove that the entire AI industry works exactly this way, but it explains well why talks of imminent collapse haven't materialized yet.

If big players are selling not only automation but also the habit of constant dialogue with machines, their economics may unfold much later—and not at all where people usually look for it.

For users, this is reason to watch not only the quality of responses but also how quickly the service becomes routine.