Behind the enthusiasm for mega-investments in artificial intelligence lies a web of financial and technological interdependencies that could transform the sector into an epicentre of global systemic risk.

In 2025, artificial intelligence became the new “digital oil”, and the giants that power it — Nvidia, Oracle, OpenAI and a handful of others — are now the custodians of the infrastructure underpinning the entire digital economy.

Yet behind the promises of unlimited growth lies a network of financial and industrial relationships so dense and circular that it more closely resembles a shadow banking system than a technology supply chain.

The golden (and hazardous) triangle of AI

In September 2025, Nvidia announced an agreement “of up to 100 billion dollars” with OpenAI to build and power 10 gigawatts of computational capacity, based on next-generation GPUs.

A month later, Oracle raised the stakes with a plan worth approximately 300 billion dollars over five years to host OpenAI’s cloud infrastructure in its own data centres, integrating Microsoft’s Azure platform into an unprecedented multi-cloud architecture.

In parallel, Nvidia will supply Oracle with hundreds of thousands of GB200 chips to power the future “Stargate” supercomputers, whilst OpenAI diversifies by signing a new agreement with AMD (6 GW of capacity and warrants of up to 10% of capital).

On paper, a synergistic ecosystem. In practice, a chain of cross-dependencies: Nvidia finances OpenAI, which buys GPUs from Nvidia, which sells infrastructure to Oracle, which in turn provides cloud capacity to OpenAI.

A seemingly perfect “golden triangle”, but one resting on extremely delicate financial and operational balances.

From competitive advantage to contagion risk

This industrial entanglement is a mine of efficiency — but also a bomb of complexity.

If any one of the three companies were to encounter an obstacle — technical, regulatory or financial — the impact would propagate along the entire chain.

  • Hardware risk: Nvidia’s chip production depends on advanced lithographic nodes from TSMC and critical materials (copper, silicon, rare earths). A production delay or a geopolitical sanction can halt entire AI computing lines.
  • Infrastructure risk: Oracle, by providing the cloud base on which OpenAI operates, concentrates in its own hands the resilience of services that have become strategically important for governments and businesses. A blackout, a vulnerability or a network congestion event would become a systemic incident.
  • Financial risk: the commitments running into hundreds of billions are structured across multiple layers of leverage and future return projections. If the real demand for AI — in terms of applications, subscriptions or enterprise services — does not grow as forecast, the pressure on balance sheets will become severe.
  • Regulatory and reputational risk: OpenAI is the public face of artificial intelligence, and any scandal linked to bias, security or improper use of data can rebound on Oracle and Nvidia as well, since they are its suppliers and strategic partners.

It is a closed ecosystem, in which the failure of one actor can generate cascade effects similar to those of a financial default: supplies slow, contracts slip, assets and share prices are written down, margin calls are triggered, and capital withdraws.

The bubble of expectations

The current rally in the AI sector, driven by the shares of Nvidia, Microsoft, Oracle and AMD, is reminiscent in some respects of the dot-com bubble of 2000.

Valuation multiples remain at levels unsustainable relative to fundamentals: Nvidia’s market capitalisation exceeds three trillion dollars, and a significant portion of its current value is discounted on still-hypothetical future revenues.

Meanwhile, energy and infrastructure costs are exploding: every gigawatt of AI capacity requires investment in the order of 8–10 billion dollars, and the electricity requirements of new data centres could surpass those of entire European nations.

AI, in short, is as much a macroeconomic wager on energy, debt and investor confidence as it is a technological revolution.

Technological systemic risk: a new digital Lehman

What emerges is a technological systemic risk: a new type of vulnerability that combines supply chains, finance and data in a single closed circuit.

The interconnection between chip suppliers, cloud platforms and model developers generates a potential domino effect that the markets, until now, have chosen to ignore.

Just as in 2008 the hidden financial leverage embedded in subprime derivatives was “invisible” until it collapsed, today the leverage of AI is technological and contractual: decade-long contracts, obligations on future capacity, investment advances contingent on the exponential growth of models.

If that growth were to slow — due to physical, regulatory or simply demand-side constraints — the edifice of projections could deflate within a few quarters, generating an “AI crash” of unpredictable proportions.

Reading the weak signal

The most astute investors are already monitoring early warnings:

  • the slowdown in demand for high-end GPUs from smaller cloud providers;
  • the first antitrust investigations into concentration in the AI market;
  • the debate over an “AI Tax” to cover the energy consumption of data centres;
  • and, above all, the growing difficulty of converting hype into sustainable profits.

Oracle, Nvidia and OpenAI are building the infrastructure of the future, but they are doing so in a context of strong mutual dependence and leveraged capital, where every unfulfilled promise can translate into a fracture of confidence.

Conclusion: the new interdependence of technological power

Artificial intelligence is not merely a frontier of innovation: it is also a new architecture of economic power.

At its core, a handful of companies control the flow of data, energy and cognitive capital.

It is an extraordinarily efficient system — and for precisely that reason, a fragile one.

When three giants such as Oracle, Nvidia and OpenAI become intertwined at this level, the risk is no longer purely industrial: it becomes macro-financial.

History teaches that every great revolution — from railways to electricity, from the internet to derivatives — has known its “moment of euphoria” before finding equilibrium.

AI will be no exception.