The Great AI Roundabout: How Circular Funding Inflates the AI Bubble

The Artificial Intelligence (AI) boom has delivered monumental shareholder value, transforming companies like Nvidia into trillion-dollar entities and fuelling the spectacular valuations of AI labs like OpenAI and Anthropic. Yet, beneath the veneer of exponential growth, a complex web of financial arrangements—dubbed the "Circular AI Economy"—is sparking concerns that these soaring valuations are being artificially inflated, potentially setting the stage for a painful market correction.

This critical analysis dissects the movement of funds between chipmakers, AI organisations, and hyperscalers, arguing that this circularity obscures true underlying commercial demand and poses systemic risks to the entire AI ecosystem.

🔄 The Mechanics of Circular Financing

The core of the issue lies in strategic cross-investments and creative financing structures that see funds effectively move from a supplier to a customer, only to return to the supplier as revenue. Nvidia, as the dominant supplier of the essential GPU hardware, sits at the centre of this mechanism.

The Nvidia-AI Lab Axis

The most striking examples involve the massive investments made by chip manufacturers into their largest customers. Consider the landmark $100 billion investment commitment made by Nvidia to OpenAI. In this arrangement, Nvidia provides capital for OpenAI's colossal infrastructure build-out. OpenAI, in turn, uses a significant portion of that funding to purchase or lease Nvidia's chips (such as the high-end H100 and B30A GPUs).

• The Circular Flow: Nvidia invests cash/debt financing -> OpenAI receives funds ->OpenAI orders Nvidia GPUs -> Nvidia records massive revenue.

  
  

While these deals secure long-term demand for Nvidia and accelerate development for OpenAI, critics see it as "revenue recycling." Nvidia boosts its topline revenue and market dominance, while OpenAI's staggering valuation (which has reached $500 billion in private markets) is sustained by capital that originates, in part, from its primary supplier. This arrangement locks in the AI lab to Nvidia's proprietary CUDA software ecosystem, creating an almost unassailable moat.

The Hyperscaler-AI Lab-Nvidia Triangle

The circularity extends to the major hyperscalers (Microsoft, Amazon, Google, etc.). These companies are simultaneously customers of Nvidia (buying chips for their cloud services), investors in AI labs (e.g., Microsoft's multi-billion-dollar stake in OpenAI), and competitors to both (developing their own custom AI chips, or ASICs).

Hyperscalers compete to attract AI labs by offering enormous packages of cloud credits—essentially free computing time.

• A startup receives a massive grant of Google or AWS credits.

• The startup uses these credits to train models, consuming the compute resources (i.e., the GPUs) provided by the hyperscaler.

• The hyperscaler had to purchase these expensive GPUs almost exclusively from Nvidia.

  
  

This system creates a triple-loop effect: Nvidia sells chips to Hyperscalers; Hyperscalers give away compute to AI Startups; the AI Startups need this compute to justify their enormous valuations in subsequent funding rounds, which encourages Hyperscalers to buy even more Nvidia chips. The entire chain is built on massive capital expenditure that hasn't yet translated into proven, durable profitability for the end-users.

⚠️ Inflated Valuations and the Bubble Concern

The primary concern about this circular funding model is its direct contribution to the inflation of valuations across the AI sector, drawing unnerving parallels to the 2000 dot-com bubble.

Distorted Revenue Quality

When a large portion of a company's revenue originates from a customer who was directly financed by the supplier, the quality of that revenue is questionable. It is not organic demand driven by end-user consumption but rather a strategically orchestrated deployment of capital to secure future market share. This can lead to misleading financial signals:

1. High Revenue, Low Profitability: AI labs like OpenAI are posting substantial revenue figures but still operate at significant losses, suggesting that the massive costs of the chips and compute (the money flowing back to Nvidia and the hyperscalers) far outpace the money generated from actual commercial use.

   
   

2. Private Market Risk: Many of the most highly-valued AI companies remain private, raising capital at escalating valuations without the rigorous financial scrutiny of public markets. These valuations are often based on projected earnings many years in the future, rather than current business fundamentals.

The Specter of Vendor Financing

Nvidia's strategic deals, including debt arrangements where GPUs themselves are used as collateral, are essentially a modern form of vendor financing. This harks back to the telecom bubble, where equipment manufacturers like Lucent funded their customers, who then bought Lucent's equipment, ultimately creating an unsustainable financial structure that collapsed when the end-user business models failed. While today's AI customers (the hyperscalers) are far more financially robust, the principle of a supplier propping up its own demand remains a substantial risk factor.

Accounting Practices

Further concerns have been raised about accounting choices, such as the depreciation schedules used for these expensive AI assets. Some critics argue that GPUs become obsolete much faster than their stated depreciation period (e.g., a six-year useful life), meaning that the true cost of compute is being underestimated. This practice, if widespread, artificially inflates the near-term reported earnings of the companies buying the hardware, further justifying their inflated valuations.

⚖️ The Critical Need for AI Tax

The debate over the Circular AI Economy underscores the urgent need for a regulatory and fiscal response, such as the implementation of an AI and Robot Tax—the core focus of Tax The AI.

If a significant portion of the AI economy's value is being generated not by sustainable, organic commercial activity but by complex financial engineering and capital deployment designed to secure a monopoly position, then the traditional tax base is being eroded. The current system rewards the hoarding of computational power and the inflation of valuations, while the societal costs of automation are borne by the public.

A Robot Tax or a Compute Tax could address this by:

1. Stabilising the Economy: By taxing the high-volume, potentially artificial transactions that fuel the circular economy, it could cool down speculative investment and incentivise more sustainable business models driven by true productivity gains, not just market dominance.

   
   

2. Recuperating Societal Costs: It ensures that the entities consolidating this immense wealth and power contribute a fair share towards mitigating the effects of automation on employment and funding the public infrastructure (like education and social security) that supports the AI workforce.

   
   

3. Promoting Transparency: A clear tax mechanism based on compute power consumption or transaction volume could provide a more objective measure of activity than the often-opaque private market valuations and circular revenue figures.

   
   

The AI Roundabout is a powerful engine of technological advancement, but the financial dynamics driving it are creating an increasingly fragile bubble. As a society, we must scrutinise the quality of this "AI value" and use fiscal policy to ensure the economic benefits of this revolution are broadly shared, rather than concentrated in a self-reinforcing financial loop.