The Circular Funding Deals Driving the AI Boom and the Looming Bubble Risk

"The wager to end them all." That is how Wall Street is pricing the current artificial intelligence boom—as a miracle that cannot fail. Yet, behind the spectacle of trillion-dollar valuations and boundless growth projections lies a precarious financial structure heavily reliant on circular deals and an infrastructure arms race that recalls the most volatile periods of modern market history.

In a recent Bloomberg Originals examination of the AI economy, reporters Shirin Ghaffary and Carmen Reinicke, alongside columnists like Dave Lee, analyzed the structural shifts underpinning the massive commitment of capital by tech giants such as Microsoft, Google, and Nvidia. Their collective assessment focuses on the financial mechanics—specifically the multi-billion-dollar circular deals—that are simultaneously fueling unprecedented infrastructure growth and raising the specter of a catastrophic economic bubble.

The core mechanism under scrutiny is the "circular deal," a process where money, products, or services are exchanged back and forth between a small cohort of interconnected companies, often creating the illusion of organic demand and robust revenue streams. Nvidia, the undisputed king of AI hardware, sits at the epicenter. Nvidia invests heavily in foundational AI companies like OpenAI and Anthropic, while those same AI developers simultaneously become Nvidia’s largest customers, purchasing the high-performance GPUs necessary for training their large language models. This creates a closed loop where investment dollars quickly return to the hardware supplier as revenue.

This financial merry-go-round is not limited to just a few entities. As Bloomberg News reporter Shirin Ghaffary illustrates, the web of capital extends to cloud providers like Oracle, which sometimes leases compute power to OpenAI. This means Oracle is a customer of Nvidia for chips, and OpenAI is a customer of Oracle for compute, while Nvidia is an investor in OpenAI. “This money is kind of spinning around the same companies,” notes Dave Lee, a columnist for Bloomberg Opinion. These arrangements, while not illegal or inherently inappropriate in principle, reach a scale where they can inflate valuations to dizzying heights, potentially masking true market performance and creating systemic fragility.

The sheer volume of capital expenditure committed to this AI race is staggering, diverting billions into tangible assets far removed from Silicon Valley’s software-centric image. This boom is not just about code; it’s about construction, energy, and water—the real-world resources required to power the immense computational demands of foundational models. According to reporting, companies are expected to spend $3 trillion on AI data centers in the coming years. This is driving an infrastructure build-out arms race, with data centers and power stations being the only construction sectors projected to see significant growth in the near term, even as spending on factories, warehouses, and commercial real estate declines.

Sam Tabar, CEO of WhiteFiber, highlights this infrastructural imperative, noting the advantage of retrofitting old industrial sites: "If you can get up and running in six months using a retrofit format versus a greenfield format, which takes up to two years, that’s a much better proposition." This rush to build out capacity quickly reflects the intense competitive pressure in the AI space, where time-to-market and access to compute are paramount. The underlying assumption driving this massive capital deployment is that AI will inevitably become profitable, yielding returns that justify the initial, enormous investment.

However, the immediate concern remains the lack of clear profitability among the most celebrated AI startups. These companies are burning through cash at an astonishing rate simply to keep their models running and to scale up. “Every time someone uses ChatGPT, OpenAI likely loses money,” Lee observes, underscoring the gap between technical prowess and sustainable economics. The path to profitability for these companies is often projected years into the future, creating a scenario where current valuations are based almost entirely on the promise of future, exponential productivity gains that have not yet materialized across the broader economy.

The financial risk is compounded by the historical precedent of the dot-com bust. While the internet boom fundamentally changed the world, the subsequent crash wiped out $5 trillion in value and took years for even the strongest companies to recover. Amazon’s stock, for instance, took eight years to return to its pre-crash peak. Cisco, a "picks and shovels" company of that era, took 25 years. The current AI boom shares a key characteristic with the dot-com era: circular dealmaking that inflated demand for infrastructure—in the 90s, it was fiber optic cable; today, it is GPUs and data center capacity.

The critical difference now, however, is the sheer scale and systemic integration of the players involved. The largest tech companies—Amazon, Microsoft, Google, Apple—are now so deeply entrenched in the global economy, and their stocks are so widely held in 401(k)s and index funds, that their failure could trigger consequences far greater than the dot-com crash or even the 2008 financial crisis. This raises the question of whether these companies, and the infrastructure providers they rely on, are becoming "too big to fail." Should the demand for AI products suddenly weaken, or should the timeline for profitability extend further than investors anticipate, the ensuing correction could create a far-reaching economic shockwave. The industry is currently engaged in "the wager to end them all," betting that the underlying utility of AI will eventually validate the astronomical valuations and the unprecedented infrastructure build-out that sustains them.