A trillion-dollar order: Where does OpenAI's money come from?

```

OpenAI has signed computing power procurement agreements totaling nearly $1 trillion, but clearly, it does not have that much money in its bank account.

This year, the company leading the global AI wave has signed computing power agreements totaling nearly $1 trillion at an unprecedented pace with giants such as Nvidia, AMD, and Oracle—a scale far exceeding its own revenue and financing capacity.

This has triggered a core question in the market: How can a startup that is still losing huge amounts of money and continually “burning cash” support this massive gamble?

The answer isn’t traditional equity or debt financing, but a carefully crafted kind of “financial alchemy”: through “circular financing” that binds suppliers closely and an innovative “equity-for-procurement” model, OpenAI is rewriting the capital rules of the AI era.

Behind the Trillion-Dollar Computing Orders: A Capital Game of “Letting Partners In”

OpenAI's trillion-dollar computing power plan consists of a series of heavyweight agreements with tech giants.

Specifically, these agreements include deals with chip giant Nvidia worth up to $500 billion, with AMD up to $300 billion, and with cloud service provider Oracle $300 billion. In addition, there’s a deal with data center group CoreWeave exceeding $22 billion, aiming to provide OpenAI with over 20 gigawatts (GW) of computing capacity over the next decade—energy consumption roughly equivalent to 20 nuclear reactors.

Behind this grand blueprint lies a harsh financial reality. OpenAI is burning cash at an astonishing rate. Gil Luria, an analyst at research firm DA Davidson, bluntly stated: “OpenAI simply does not have the ability to make any of these commitments,” and predicts the company might lose as much as $10 billion this year.

So, where does the money come from?

Media analysis says the answer lies in OpenAI’s clever use of the Silicon Valley “fake it until you make it” philosophy, employing so-called “circular arrangements” to get a slew of tech giants to “have skin in the game” for its future.

In other words, this is essentially a capital game: OpenAI uses its leading position and massive market expectations in AI as leverage, persuading suppliers to fund its computing expansion in non-traditional ways, transforming future growth potential into current purchasing power.

“Circular Revenue” and “Equity-for-Procurement”: The Financial Alchemy Behind the Trillion-Dollar Orders

OpenAI’s financial magic is mainly realized through two innovative models with AMD and Nvidia—different, but logically similar. These models clearly show how AI computing resources can evolve from a mere capital expenditure into a financialized, securitized asset class.

(1) The AMD Model: “Equity-for-Procurement” Based on Future Performance

The agreement with AMD is unprecedented in semiconductor history.

On the surface, OpenAI plans to purchase and deploy up to 6GW of AMD Instinct GPUs worth $90 billion. But at the core of the agreement is that AMD issued warrants to OpenAI, allowing it to buy up to 160 million AMD shares at a nominal price of just $0.01 per share.

This structure is essentially a financial instrument that cleverly converts hardware sales into equity allocation. If, as calculated, AMD's stock price rises to $600 thanks to large-scale adoption of its GPUs, the potential value of OpenAI’s shares would reach $96 billion—roughly equal to the hardware purchase total. This means, if the collaboration goes smoothly and the market is optimistic, OpenAI could almost “acquire this computing power for free.”

The ingenuity of this model lies in it being a performance-based equity incentive. It directly ties AMD’s long-term valuation to the growth of OpenAI’s infrastructure, forging a powerful symbiosis and providing OpenAI a potential “self-funding” path: as procurement milestones are met and AMD’s share price rises, OpenAI can cash out exercised shares to finance further GPU purchases, greatly alleviating its capital expenditure pressure.

(2) The Nvidia Model: More Direct “Circular Revenue”

In contrast, cooperation with Nvidia uses a more direct circular logic. Nvidia plans to invest up to $100 billion in OpenAI over the next decade. This capital can be directly used by OpenAI to purchase Nvidia chips.

Goldman Sachs’ deep analysis report calls such transactions “circular revenue”: Nvidia injects funds into OpenAI, and OpenAI then places GPU orders with Nvidia using this money.

The firm believes that in this model, the client isn’t really ”self-funding”; when the supplier’s equity investment is ultimately recycled as revenue, investors look more strictly at the “circular” nature of that revenue, which could lead the company to take a “more cautious” stance on valuation multiples for key partners like Nvidia.

Goldman Sachs Analyzes OpenAI’s Cash Flows

Although the financial instrument design is intricate, OpenAI may still face a huge funding gap.

To more accurately understand OpenAI’s finances, Goldman Sachs analyzed it from two angles:

First, from the operating cash flow perspective, things seem manageable. Goldman Sachs estimates that OpenAI’s annual operating infrastructure costs (including inference, training, personnel, etc.) in 2026 will be about $35 billion. Sources of funds are approximately:

OpenAI’s own revenue: about 48%.
Vendor financing: about 27%.
External equity/debt financing: about 25%.

In this model, OpenAI seems to be able to cover most daily operating expenses through its own cash generation and partner support, with a relatively balanced capital structure.

However, once major future capital commitments are considered, a very different picture emerges.

Goldman Sachs notes that aside from operating costs, OpenAI has made major outside capital commitments to secure future computing power. These include: collaborating with Nvidia on building its own data centers (requiring about $60 billion capital in 2026), and funding the “Stargate” project (estimated to require $19 billion in 2026).

Factoring in these huge future capital expenditures, Goldman Sachs calculates OpenAI's total funding needs in 2026 would skyrocket from $35 billion to about $114 billion. Faced with this huge gap, the funding structure changes dramatically:

External equity/debt financing: reliance surges to an overwhelming 75%.
OpenAI’s own revenue: contribution is diluted to 17%.
Vendor financing: drops sharply to 8%.

In this model, the funding structure is utterly unbalanced. External equity and debt financing needs rise to 75%, while self-generated revenue is diluted to just 17%.

In other words, even with various innovative financial instruments, to actually fulfill major capital commitments, OpenAI’s reliance on external capital remains staggering. Its grand computing power empire cannot be sustained by its business or clever supplier agreements alone; the ability to continually secure large-scale financing from VCs and debt markets is the key to success in this massive gamble.

High Stakes, Risks, and the Future: AI Bubble or Industry Cornerstone?

Whether it’s AMD’s equity incentives or Nvidia’s direct investment, these agreements instantly boosted partners’ market cap. For example, after announcing cooperation with OpenAI, Oracle’s and AMD’s market caps soared by $244 billion and $63 billion, respectively.

This rise in share prices, in turn, further boosts market confidence in such cooperation models, forming what appears to be a perfect positive feedback loop, but at the same time ringing alarm bells about an “AI bubble.”

For now, OpenAI’s complex financial architecture might solve the source of computing power in the short term, but the risks should not be underestimated:

Credit and growth risks: Ratings agency Moody’s has issued a warning to Oracle because its future data center business is overly reliant on OpenAI—a client that has yet to prove its profitability. The ecosystem’s prosperity is built on the assumption that AI applications will continue exponential growth. If user growth slows or willingness to pay decreases, this high-expectation capital chain could break.
Lack of cost discipline: A veteran Silicon Valley investor commented that OpenAI “inherently has no cost discipline,” comparing it to the early days of Amazon and Oracle, both of which only started strict cost control after being on the verge of bankruptcy. CEO Sam Altman has publicly said profitability “is not among my top ten concerns.” Such a growth-first strategy works when capital is plentiful, but could face huge pressure if the market shifts.
Deep shifts in industry structure: Goldman's report notes this trend is changing the customer structure of upstream companies like Nvidia. Its revenues are shifting from traditional, financially strong “hyperscalers” to AI startups and sovereign AI funds, which rely more on external financing and are riskier. By 2027, Goldman predicts AI startups and sovereign AI will contribute $66 billion and $46 billion of Nvidia’s revenue, respectively. This change brings new growth points but also greater volatility and valuation uncertainty to the whole industry chain.

Epilogue

OpenAI’s trillion-dollar computing contracts are a capital maneuver leveraging financial innovation, industrial alliances, and market expectations to the utmost. Through “circular financing,” it has leveraged resources many times its own size, accelerating AI infrastructure construction.

However, the foundation of this system is a firm belief in continuous breakthroughs in technology and sustained market fervor.

Is this laying a solid computing foundation for the next industrial revolution, or is it a castle in the air built on circular credit and a capital bubble? Time will tell. But one thing is certain: OpenAI has brought itself and its partners to the table of an unprecedented high-stakes gamble.

This article is from WeChat public account “Ying AI”. For more AI news, click here.

Risk Reminder and DisclaimerThe market has risks, investment needs caution. This article does not constitute personal investment advice, nor does it take into account individual users’ specific investment objectives, financial circumstances, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable to their particular situation. You are solely responsible for any investment made accordingly. ```