OpenAI's financial disclosures reveal a stark reality: the company is hemorrhaging money at an unsustainable rate, with a $20.9 billion operating loss in 2025 despite generating $13 billion in revenue. This gap between revenue and spending exposes a fundamental challenge facing the entire AI industry as companies like OpenAI and Anthropic prepare for public offerings.

How Bad Are OpenAI's Losses Really?

The numbers tell a sobering story. OpenAI spent roughly $34 billion across four major categories in 2025: $19.2 billion on research and development (training costs), $7.5 billion on inference costs (running the models), $5.7 billion on sales and marketing, and $1.6 billion on general administrative expenses. To put this in perspective, OpenAI's annual spending roughly equals what the U.S. government spent on the FBI and NASA combined in a single year.

The company's $5.7 billion marketing budget alone exceeded the entire world's spending on podcast advertising for the year. With that budget, OpenAI could have purchased every Super Bowl ad slot for the past seven years. This raises a critical question: how much of this spending is truly necessary to build and operate an AI company?

Some have suggested OpenAI's true loss was closer to $8 billion when excluding one-time charges related to stock compensation and the company's reorganization into a for-profit entity. However, this argument doesn't hold up under scrutiny. Operating loss, which excludes capital expenditures and one-time charges, still reached $20.9 billion. This metric is harder to manipulate with accounting adjustments, making it a more reliable measure of whether the business actually works.

Is Anthropic in Better Shape?

Anthropic presents a somewhat less dire picture, though the company remains deeply unprofitable. The company reported $4.5 billion in revenue for 2025, with annual recurring revenue hitting $47 billion by May 2026. However, the critical question remains unanswered: how much does Anthropic actually spend to generate that revenue ?

Wall Street Journal estimates pegged Anthropic's compute costs at $7 billion in 2025, but the same publication underestimated OpenAI's compute costs by roughly 60 percent. If the Journal made a similar error with Anthropic, the company's actual compute spending likely reached around $11 billion. Adding estimated costs for sales, marketing, and administrative expenses suggests an operating loss of approximately $11 billion, roughly half OpenAI's losses but still unsustainable.

Recent reports claimed Anthropic was on track to achieve profitability in the second quarter of 2026. However, this projection relied on a two-month window when a major SpaceX contract was heavily discounted. Once that discount expired, the contract would balloon to $15 billion annually, making the profitability claim largely theoretical.

What Would It Take for AI Companies to Become Profitable?

For Anthropic to turn profitable this year, two conditions would need to align: the company would need to book the full $47 billion in annual recurring revenue it projects, and it would need to limit cost growth to no more than 150 percent. Both scenarios are described as extremely difficult.

OpenAI faces an even steeper climb. Even if the company flatlines spending growth, revenues would need to grow exponentially to offset the $20.9 billion operating loss. This scenario appears increasingly unlikely, especially as OpenAI considers marking down ChatGPT prices to compete in the market.

Steps AI Companies Must Take to Reach Profitability

• Ruthless Cost Elimination: Identify which expenses are truly necessary for operations and aggressively cut the rest. This means moving beyond experimental projects and focusing on core business functions that generate revenue.
• Marketing Efficiency: Reduce bloated sales and marketing budgets that exceed industry norms. OpenAI's $5.7 billion marketing spend suggests the company is trying to force adoption rather than letting product quality drive growth organically.
• Revenue Acceleration: Grow revenue faster than costs. This requires either raising prices, expanding customer bases, or developing new revenue streams that don't require proportional increases in compute spending.

Why This Matters for Sam Altman and the AI Industry

The profitability crisis extends beyond OpenAI and Anthropic. Anthropic lost nearly $6 billion in 2024, while xAI lost $2.5 billion in a single quarter. These losses are not sustainable indefinitely, particularly as these companies seek public market valuations.

The fundamental problem is that AI companies have been operating like high-growth startups with venture capital backing, spending aggressively on research and marketing without regard for profitability. As these companies prepare for public offerings, investors will demand a clear path to sustainable profits. Right now, that path doesn't exist.

Some analysts suggest AI companies could follow SpaceX's model: maintain a small public float, operate in a massive addressable market, and rely on a charismatic CEO to maintain investor confidence. However, this strategy requires a personality-driven leadership style that neither Sam Altman nor Dario Amodei (Anthropic's CEO) have demonstrated.

The most critical hire at these companies is no longer the CEO, but the CFO. AI labs must transition from the mindset of building cutting-edge technology at any cost to the discipline of building profitable businesses. Without this shift, even companies with revolutionary technology will struggle to justify their valuations to public market investors.

As one analyst concluded, if AI can't make more money than it spends, it's uninvestable. The technology may be transformative, but the business model must work.