OpenAI faces the challenge of an still fragile business model

Twelve zero figures circulate around OpenAI. Hundreds of billions are talked about like the weather. But these figures are not profits. They are bets. Future commitments. Debts maybe. Behind the glitter of artificial intelligence, a question emerges: Can OpenAI still keep the financial pace it has imposed on itself? Or is it running toward a data wall, pockets empty?

The cloud ogre or the devourer of billions

OpenAI has partnered with cloud giants. $250 billion at Microsoft , 38 at Amazon. Total: $288 billion in contracts for 36 gigawatts of computing power. Barely a third will be active by 2030. The rest? To be paid, nonetheless. According to HSBC, the bill would reach $792 billion in 2030, and up to $1.4 trillion in 2033 if the pace continues.

Bank analysts do not mince words :

Given the closely intertwined relationships between AI LLM companies, cloud, and semiconductor companies, we believe there is an argument for some flexibility, at least on the part of the largest players… less capacity will always be better than a liquidity crisis.

If Sam Altman wanted a sovereign AI, he tied it to cloud empires. And these empires do not extend credit forever. Each payment cycle looks more and more like a fight not to fall into logistical bankruptcy.

Credit growth: OpenAI’s big stretch

OpenAI aims high: 3 billion users by 2030. About 10% paying subscribers. That represents 220 million payers on ChatGPT — compared to 35 million today. That would be one of the largest subscription services in the world, ahead of Spotify, almost at Netflix’s level.

Even with a projected free cash flow of $282 billion, asset sales, and an estimated treasury of $17.5 billion in 2025, it does not add up. HSBC estimates an additional $207 billion funding need by 2030. That, despite all efforts to diversify revenue streams or strengthen direct monetization.

The contrast is striking: the company is perceived as a unicorn generating gold at every prompt, but its financial flows tell another story. HSBC, straightforwardly, describes it as a cash sink, not a profit machine. Even in its optimistic scenarios, the firm acknowledges that OpenAI will continue to subsidize the vast majority of its users for years to come.

In the absence of a massive fundraising — or spectacular growth — OpenAI risks continuing to massively subsidize its services. Each dollar raised risks going straight into the pockets of cloud providers. The dream of “AI for all” could turn into a budget runaway, supported more by investors’ faith than by actual revenues.

OpenAI and the myth of 220 million AI subscribers

The strategy is clear: monetize artificial intelligence through subscriptions. The Plus ($20/month) and Pro ($200/month) plans pay off. But 5% conversion today, 8.5% expected by 2030? The gap is wide. Hope also lies in targeted ads, assisted shopping, or commission interfaces.

However, OpenAI is not alone. Anthropic, xAI, and others are gaining momentum. OpenAI’s AI market share, dominant today, could slip from 71% to 56% in the consumer sector, and from 50% to 37% in the enterprise sector.

And usage? It’s exploding… but not necessarily revenues. The free base grows faster than payers. “AI for All” hits the freemium rule: the more people you have, the more you pay to serve them computing, with no immediate return.

Key data to remember

• $288 billion in cloud contracts signed;
• 36 GW of engaged power;
• $207 billion still to raise by 2030;
• 220 million targeted subscribers, 35 million current;
• OpenAI AI market share expected to decline.

By 2026, OpenAI hopes for an IPO valued at $1 trillion . This IPO would be a poker shot to refinance the global ambition, while imposing artificial intelligence as the technological pillar of the century. But will this raise be enough to justify all the zeros already engaged? Nothing is less certain in a world where speed sometimes exceeds meaning.