A leaked financial presentation, reported by the Financial Times on September 18 and cited by Bloomberg and other outlets, gives the clearest picture yet of what it actually costs to run the world’s most watched AI company. The numbers are staggering, and for businesses building on top of AI services, they raise questions worth thinking about carefully.

The Numbers

According to the leaked July 2026 document, prepared in connection with a major computing deal, OpenAI projects:

• Negative free cash flow of $278 billion from 2026 through the end of 2030
• Revenue growing from $36 billion this year to $350 billion by 2030
• Compute and infrastructure spending of approximately $856 billion by end of 2030
• The $122 billion OpenAI raised in March 2026 is expected to last only until 2028

The projections show improving cash flow trends. An earlier May 2026 internal estimate had pegged negative FCF at $305 billion, so the new July figure represents roughly $27 billion in improvement. Much of the infrastructure build is also financed by computing partners rather than drawn directly from OpenAI’s balance sheet, which partially explains how spending can rise while the cash burn outlook improves.

Still, the headline is hard to ignore: OpenAI does not project becoming profitable before 2030, even as it targets revenue that would make it one of the largest software companies in history.

Why These Numbers Matter to Business Owners

Most business owners aren’t watching OpenAI’s financials. They probably should be, at least at a headline level, because these numbers reveal something important about the economics of frontier AI.

Frontier AI is extraordinarily expensive to run. The compute costs alone are roughly 2.4x OpenAI’s projected revenue over the same period. That gap isn’t a sign of mismanagement. It reflects the actual cost of training and running the most capable models in the world. Data centres, GPU clusters, energy, and the people who operate them don’t come cheap.

Pricing pressure is real and structural. AI service prices have dropped dramatically in the past two years as competition intensified. That’s good for buyers in the short term. But when the provider is burning through hundreds of billions more than it earns, the long-run pricing picture is less certain. Rates that look cheap today are being subsidised by capital markets, not sustainable unit economics.

Dependency risk deserves attention. If a significant part of your business workflow depends on a single AI service, understanding that service’s financial stability isn’t paranoia. It’s basic vendor risk management. OpenAI is raising capital actively and has committed backers, but the scale of spend it’s projecting means it needs those capital markets to keep working.

What This Means in Practice

None of this means businesses should stop using AI tools. The productivity gains from AI-assisted work are real and documented. But it does mean a few things are worth taking seriously:

Right-sizing matters. Not every workflow needs frontier model capability. Open-weight models, smaller fine-tuned models, and purpose-built AI tools can deliver most of the value at a fraction of the cost and with fewer dependency risks. Businesses that have mapped their AI use cases clearly are better positioned to use the right tool for each job rather than defaulting to the most expensive option.

Don’t build on a single point of failure. An AI strategy that runs entirely through one provider’s API is a concentrated bet. Thoughtful deployments consider fallback options, model interoperability, and what happens if pricing or availability changes.

The infrastructure buildout isn’t slowing down. OpenAI’s $856 billion compute commitment, whatever its financing structure, signals that the major players believe the demand for AI capability is going to continue growing. For businesses, that means AI’s usefulness in real operations is only going to expand. The question isn’t whether to engage with AI. It’s how to engage with it strategically.

The Broader Picture

OpenAI isn’t alone in this pattern. Every major AI lab is running at a loss relative to the scale of its compute commitment. Google, Microsoft, Amazon, and Meta are all spending at a scale that only makes sense if the long-run demand for AI proves as large as they believe it will.

The bet these companies are making is that AI will become infrastructure as fundamental as cloud computing, and that the returns will compound over decades, not quarters. That bet may well be right. But it’s a bet, and it’s being funded by capital markets at scale.

For Enterprise DNA’s audience of business owners and data professionals, the practical message is straightforward: use AI aggressively where it creates genuine value, build that use on a foundation that isn’t wholly dependent on any single provider’s survival, and pay attention to what you’re actually getting for what you’re spending. The vendors betting their existence on this technology have their reasons. Make sure yours are your own.

---

Sources: Financial Times (paywalled, September 18, 2026), as reported by Bloomberg. Additional context from The Next Web and multiple financial outlets.