OpenAI's $665 Billion Gamble: The Astronomical Cost of the AGI Race

OpenAI has once again recalibrated its financial telescope, and the view is becoming increasingly expensive. The company has revised its cumulative cash burn forecast through 2030 to a staggering $665 billion, marking an increase of roughly $111 billion over previous estimates. This upward adjustment, driven by the spiral of inference and training costs, underscores the brutal economic reality of the race toward Artificial General Intelligence (AGI).

For the AI industry, this serves as a stark wake-up call. While revenue is climbing—OpenAI reportedly generated $13.1 billion in 2025, tripling its previous year's performance—the cost of operating frontier models is growing at an even faster clip. The narrative has shifted from pure growth to a high-stakes battle for capital efficiency, where the "burn now, profit later" strategy is being tested to its absolute limits.

The New Financial Reality

According to internal documents cited by The Information, the revision stems from a combination of higher-than-expected compute costs and the sheer scale of infrastructure required to train next-generation models. The company does not expect to become cash-flow positive until 2030, a timeline that pushes profitability further into the future than many investors might have anticipated.

The updated projections paint a picture of massive capital consumption over the next five years. To put the $665 billion figure into perspective, it exceeds the GDP of many mid-sized nations and dwarfs the capital expenditure of traditional tech giants in their early years.

Breakdown of Projected Cash Burn

The following table outlines the dramatic shift in OpenAI's financial expectations, highlighting the deepening deficit before the projected turnaround in 2030.

Data derived from internal financial projections reported by The Information. Figures represent annual cash burn/flow.

The Inference Trap and Margin Squeeze

A critical driver of this financial recalibration is the cost of inference—the computing power required to run the models every time a user sends a query. In 2025, OpenAI’s inference costs reportedly quadrupled. This surge is a double-edged sword: it indicates massive user engagement, with Weekly Active Users (WAU) hitting 910 million, but it also erodes profitability for every interaction.

The impact on margins has been severe. OpenAI’s adjusted gross margin plummeted to 33% in 2025, falling significantly short of its 46% target. For context, typical software-as-a-service (SaaS) companies often boast margins exceeding 70%. The company has subsequently adjusted its long-term margin goals, now aiming for 52% to 67% by the end of the decade—a clear admission that the "software economics" investors love may not fully apply to foundation model providers in the near term.

Training costs are equally daunting. The company projects spending nearly $440 billion on model training alone through 2030. This includes $32 billion in 2026 and $65 billion in 2027, flowing largely into the coffers of partners like Microsoft, Oracle, and NVIDIA for cloud capacity and GPUs.

Revenue Streams: Beyond Subscriptions

Despite the cash burn, OpenAI’s revenue engine is firing on all cylinders. The consumer division remains the crown jewel, projected to generate $150 billion by 2030. However, the company is aggressively diversifying:

• Enterprise Growth: B2B revenue is forecast to leap from $2 billion in 2025 to $70 billion by 2030.
• API Revenue: Expected to contribute $47.5 billion.
• Hardware Ambitions: OpenAI is betting on new form factors, anticipating its first hardware revenue in 2026 ($100 million), scaling to $15 billion by 2030.

This diversification is essential. Reliance solely on $20/month subscriptions is insufficient to cover the capital expenditure required for clusters of hundreds of thousands of H100 (and future generation) GPUs.

The Competitive Landscape: Anthropic's Leaner Path

The revised forecast places OpenAI in a precarious position relative to its competitors. Anthropic, its primary rival founded by former OpenAI researchers, is reportedly targeting a breakeven point as early as 2028. If Anthropic can achieve comparable model performance with a more sustainable cost structure, it could challenge OpenAI’s dominance not just in technology, but in investment attractiveness.

OpenAI is currently negotiating a funding round exceeding $100 billion at a valuation of roughly $750 billion. While backers like SoftBank and Microsoft appear committed, the extended timeline to profitability adds pressure on Sam Altman and his team to deliver "God-level" AI capabilities that justify the expense.

Conclusion: The High-Stakes Path to 2030

OpenAI’s decision to raise its cash burn forecast by $111 billion is a declaration of intent. It signals that the company views the current era not as a time to consolidate, but as a window to aggressively capture the future of computing, regardless of the cost.

For the broader AI ecosystem, this raises fundamental questions about sustainability. If the market leader requires nearly three-quarters of a trillion dollars to reach positive cash flow, the barrier to entry for new foundational model companies has arguably become insurmountable. The industry is witnessing a consolidation where only those with access to sovereign-level capital can afford to stay in the game.

As we look toward 2030, the success of this gamble will depend on two factors: whether Artificial General Intelligence can be achieved, and whether it can generate value faster than the furnaces of inference can burn through cash.