Financial Leak: OpenAI's $38.5B Deficit and the Cost of Compute

OpenAI Financial Illustration

Recent audited financial disclosures, first analyzed by Ed Zitron and subsequently confirmed by the Financial Times, have stripped away the ambiguity surrounding OpenAI's spending. The documents reveal that in 2025, the company faced a net loss of $38.53 billion** attributable to the entity, despite generating **$ 13.07 billion in revenue.

The Core Conflict: Growth vs. Burn

The central tension for CEO Sam Altman is whether his revenue-generating engine can accelerate faster than the staggering cost of compute before the company's IPO window closes. While OpenAI is ~~slowly becoming profitable~~ scaling rapidly, the "compute burn" is immense.

Financial Comparison: 2024 vs. 2025

The following table summarizes the shift in OpenAI's financial position based on the leaked documents:

Metric	2024 (Reported)	2025 (Leaked)	Change
Total Revenue	$3.7 Billion	$13.07 Billion	$\approx 3.5\times$
Total Costs & Expenses	$12.48 Billion	$34.00 Billion	$\approx 2.7\times$
Operating Loss	$8.78 Billion	$20.92 Billion	$\approx 2.4\times$
Net Loss (Attributable)	Not Specified	$38.53 Billion	N/A

Using LaTeX to express the growth dynamics: $\text{Revenue Growth} \approx \frac{13.07}{3.7} \approx 3.52$ $\text{Operating Loss Expansion} \approx \frac{20.92}{8.78} \approx 2.38$

Deconstructing the "Net Loss"

It is important to distinguish between the net loss and the operating loss. The headline figure of $38.53 billion (which stems from a total net loss of$ 60.35 billion before noncontrolling interests) is heavily influenced by accounting maneuvers.

The Accounting Nuance: A significant portion of the deficit—specifically $41.55 billion—is tied to changes in the fair value of warrant liability and convertible interests.

This means the massive net loss is partly a result of the company's complex capital structure rather than just the cost of running servers. However, the $20.92 billion operating loss remains the most critical indicator of the business's actual day-to-day economic health.

2025 Expense Breakdown

{
  "expenses_2025": {
    "research_and_development": "$19.18B",
    "cost_of_revenue": "$7.50B",
    "sales_and_marketing": "$5.73B",
    "general_and_administrative": "$1.57B"
  }
}

The "Compute as a Weapon" Strategy

Altman's strategy is clear: use massive amounts of capital to break through infrastructure constraints. By spending aggressively on compute, OpenAI aims to make its models more reliable and capable, thereby capturing the market.

Funding Milestones:

February 2026: Raised $110 billion at a$ $110 bi l l i o na t a$ 730 billion pre-money valuation.
- Contributors: SoftBank ( $30B), Nvidia ($ 30B), Amazon ($50B).
March 2026: Closed $122 billion in committed capital.
- Post-money Valuation: $852 billion.
- Lead Investors: Amazon, Nvidia, SoftBank, and Microsoft.

Current Scale Metrics:

Monthly Revenue: $2 billion
Weekly Active Users (ChatGPT): 900 million+
Paid Subscribers: 50 million+

The Microsoft Dependency

Microsoft is far more than a passive investor; it is a critical infrastructure dependency. The leaked data highlights a symbiotic, yet expensive, relationship.

Payments to Microsoft in 2025 ($17.2 Billion Total):

R&D: $10.59 billion
Cost of Revenue: $6.047 billion
Sales & Marketing: $527 million
G&A: $42 million

Additionally, OpenAI held $3.64 billion in liabilities to Microsoft at the end of the year.

Ownership Structure

Corporate Evolution & The Path to IPO

OpenAI has undergone a radical structural transformation to accommodate the astronomical costs of frontier-model development.

2015: Founded as a non-profit.
2019: Created a for-profit subsidiary to scale deployment.
Oct 2025: Restructured into the OpenAI Foundation (non-profit) and OpenAI Group PBC (Public Benefit Corporation).
June 2026: Confidentially submitted draft S-1 to the SEC.

The transition to a PBC reflects the reality that conventional startup funding is insufficient for the scale of AI research. The company is now positioning itself for the public markets, hoping that its massive user base and revenue growth will eventually outweigh the "compute burn."