Математика OpenAI: как заработать миллиарды и остаться в глубоком минусе

Sam Altman is looking for money again, and this has long ceased being news and has become more of a familiar backdrop in the industry. The situation surrounding OpenAI's finances reminds one of an old joke about a businessman who buys eggs for a ruble, boils them, and sells them for a ruble, claiming that at least he's always busy. In reality, everything is both more complicated and far more interesting.

If you carefully examine the company's financial flows, a surprising fact emerges: operationally, OpenAI is quite viable. Every model they release today generates enough revenue from subscriptions and APIs to fully cover their computing expenses and even their own research. The only problem is that in the world of generative AI, you cannot simply release a successful product and peacefully collect profits.

The industry operates in a mode of perpetual and extremely expensive maintenance. As soon as GPT-4 started generating stable cash, all that money instantly shifted to the budget for training GPT-5. And when these funds naturally ran short, they had to go begging to Microsoft and major venture funds again.

This is a classic scaling trap, taken to its absolute extreme. You must run at full speed just to stay in place, and constantly increase your pace at the same time. Silicon Valley is accustomed to unprofitable startups that burn through capital for years to capture market share.

Amazon did this for decades, Uber taught us that profit is something from history textbooks. But OpenAI has a nuance: their main asset is not a loyal user base that is easy to monetize, but computational power that becomes more expensive with each new generation of chips.

Investors are currently making a colossal bet that someday a plateau will arrive. That very blessed moment when the next iteration of the model becomes so perfect that improving it won't require exponential growth in hardware and electricity costs. But for now, the scaling law dictates the opposite logic: to achieve modest quality improvements, you need to invest ten times more resources than the last time. This turns the business model into a kind of technological pyramid, where the prosperity of today rests exclusively on hope for a miracle from the next version. If GPT-5 doesn't make a qualitative leap that justifies tens of billions of dollars in spending, even Altman's most devoted supporters will start asking questions.

Meanwhile, the company's revenues are growing at a pace that would make any software giant envious. ChatGPT Plus subscriptions and corporate contracts generate enormous sums. However, these figures look almost comical against the bills from Nvidia and plans to build data centers whose energy consumption is already measured in gigawatts.

Altman is building a future where AI becomes the new oil, but for now he more resembles an oil rig owner who spends all extracted oil just to drill the well even deeper. The dependence on Microsoft plays a key role here: the Redmond giant essentially issues OpenAI "vouchers" for its cloud computing resources in exchange for a stake in the company, creating a closed ecosystem where real money is just one element of the equation.

The risk here lies not in investors running out of money — there are still plenty willing to touch the "creators of the future." The main danger lurks in a possible intellectual dead end. If it turns out that simply adding graphics cards and internet texts no longer produces magic, the entire financial structure of OpenAI will collapse inward.

We are observing the most expensive experiment in human history, where the grand prize is creating a full-fledged artificial intelligence, and the price of losing is the bankruptcy of the most discussed company of the decade. For now, Sam Altman successfully sells hope, but mathematics is a stubborn thing, and sooner or later models will have to learn to earn faster than they manage to become obsolete.

The bottom line: OpenAI's business model is a race against its own shadow. As long as each new development requires more resources than the previous one delivered, the company remains a hostage to the market's faith in an inevitable technological miracle.