You may have seen graphics depicting the circular flow of capital between major AI infrastructure players. The interconnected spending patterns between Oracle, NVIDIA, AMD, and OpenAI have raised questions about sustainability and systemic risk in the AI sector.
The Players
Oracle is a major cloud-computing company that owns data centers full of powerful computers. These facilities are rented by other companies to run their software and applications at scale.
NVIDIA and AMD manufacture the high-end computer chips (GPUs) that power artificial intelligence workloads. These chips serve as the computational “engines” that enable computers to train and run complex AI models.
OpenAI builds and operates large-scale AI models—including ChatGPT—and requires massive amounts of computing power to develop, train, and serve these systems to millions of users.
The Circular Flow
The relationships between these companies form an interconnected loop of capital and commitments:
Oracle to NVIDIA: Oracle is spending tens of billions of dollars purchasing GPUs from NVIDIA (and increasingly from AMD) to build large-scale AI data centers. These purchases represent significant revenue streams for the chip manufacturers.
OpenAI to Oracle: OpenAI signs long-term contracts with Oracle, committing to rent computing power for several years. This spending becomes Oracle’s cloud infrastructure revenue.
OpenAI to AMD: Simultaneously, OpenAI negotiates direct deals with AMD to purchase chips, including special warrants that provide rights to acquire equity stakes. AMD secures future business, while OpenAI locks in critical supply.
The Scale of Commitments
The magnitude of these arrangements is substantial:
-
Oracle agreed to purchase approximately $40 billion worth of NVIDIA’s advanced AI chips to power OpenAI’s data center operations in Texas.
-
NVIDIA plans to invest up to $100 billion in OpenAI, tied to the development of up to 10 gigawatts of NVIDIA-powered systems over the coming years. For context, Los Angeles’s municipal utility maintains generating capacity of roughly 8.1 gigawatts.
-
OpenAI finalized a multi-year agreement with AMD, where AMD will supply AI chips totaling up to 6 gigawatts over multiple years. In exchange, OpenAI received warrants to purchase up to 160 million AMD shares—approximately 10% of the company.
Why It’s Called Circular
This arrangement is characterized as circular because capital flows remain largely within the same group of companies. Each participant’s revenue depends on another’s spending commitments, and the entire system ultimately relies on a critical assumption: that end-users will generate sufficient revenue from AI tools to support the infrastructure investments.
If actual market demand fails to materialize at the anticipated scale, the loop begins to unwind.
How The Circle Could Break
Several scenarios could trigger a breakdown in this circular system:
Revenue Shortfalls
If OpenAI fails to generate sufficient revenue from its user base, it cannot sustain the substantial bills owed to Oracle for computing infrastructure. This would leave Oracle with massive data centers containing underutilized GPU capacity that cannot generate adequate returns.
Demand Collapse
As Oracle and similar providers reduce expansion plans, NVIDIA and AMD would experience declining orders for new chips. While these manufacturers would have already recognized revenue from initial purchase waves, demand for additional capacity would evaporate rapidly.
Contract Obligations
Much of this spending is structured around long-term contracts and significant financial commitments. Companies could find themselves bound to expensive obligations even as utilization rates decline. In such a scenario, profit margins would compress quickly, equity valuations could fall sharply, and the AI infrastructure boom could transform into an overbuilt cycle.
Historical Parallels
This pattern bears resemblance to the fiber-optic network buildout of the early 2000s, when telecommunications companies invested billions in infrastructure based on projected demand that never fully materialized. The result was widespread overcapacity, financial distress, and a prolonged period of industry consolidation.
The Critical Question
The sustainability of this system hinges on a single variable: whether consumer and enterprise demand for AI applications will scale quickly enough to justify the infrastructure investments being made today. If the answer is yes, these arrangements represent strategic positioning for future growth. If not, participants face the risk of significant capital misallocation and financial losses.
The coming years will reveal whether this circular flow represents visionary infrastructure development or an unsustainable cycle of interconnected obligations.