The New Vassals: How AI Forged a New Feudal Order on Corporate Secrets

In the gleaming boardrooms of global finance and industry, a disquieting truth is setting in. The titans of the modern economy, powerhouses like JP Morgan, Citi, and Siemens, are accustomed to a world where capital is king. Yet, as the artificial intelligence revolution gathers pace, these corporate sovereigns are discovering the limits of their power. They are facing a new technological order that money alone cannot command, forcing them into a state of unprecedented dependency. The age of the corporate empire is giving way to the age of AI feudalism.

At the heart of this power shift is a stark reality: despite pouring hundreds of millions into research and development, no traditional corporation can independently build or control a state-of-the-art large language model (LLM). The cutting-edge intelligence of systems like OpenAI’s GPT-4, Google’s Gemini 2.5 Pro, or Anthropic’s Claude Opus remains the exclusive domain of a new, high-tech aristocracy. The illusion of corporate autonomy is crumbling against three fundamental barriers.

First is the matter of access. The most powerful AI models are not for sale; they are closely guarded secrets. Their “weights” the complex configuration of parameters that constitute their intelligence, are the 21st century’s equivalent of a state secret. While open-source alternatives like Meta’s LLaMA offer a taste of this power, they consistently lag behind their closed-source brethren in the sophisticated reasoning and seamless integration that enterprises crave.

Second, and perhaps more critical, is the brutal scarcity of computing power. Training a frontier AI model is an undertaking of biblical proportions, requiring armadas of specialized chips, primarily NVIDIA’s H100/H200 GPUs. The demand for this hardware so vastly outstrips supply that its creator, Jensen Huang, has effectively become the kingmaker of the AI age. He and his company act as gatekeepers, allocating this precious resource not to the highest bidder, but to strategic partners, fellow tech giants, and sovereign governments.

Finally, there is the sheer physical challenge of infrastructure. The data centers required to run these massive AI workloads are not mere server farms; they are monuments of modern engineering. They demand colossal amounts of electricity, sophisticated thermal management to prevent meltdown, and a direct pipeline to the latest chips. For most companies, building such a facility is a multi-year distraction they cannot afford.

And yet, the pressure to join the race is existential. The fear of missing out is palpable in every executive suite. Early reports show staggering productivity gains, creating a strategic risk for inaction that is seen as far greater than the risk of engagement. This has led to a great and silent surrender. Compelled to act, corporations are integrating third-party AI into the very heart of their operations. In practice, this means their employees are now actively encouraged to offload their work onto these external platforms. On a daily basis, a torrent of sensitive data- drafts of legal arguments, confidential financial models, and proprietary marketing strategies- is fed directly into a third-party intelligence that no one in the company truly controls, sacrificing data security for a competitive edge.

This dynamic has birthed a new hierarchical world order, one with a distinctly feudal structure. At the apex sit the AI Lords, OpenAI, Google, and Anthropic, who own the models and dictate the terms. Just below them is the GPU Nobility, helmed by NVIDIA, which controls the physical means of AI production.

And then there are the Corporate Vassals. These once-sovereign entities now rent access to intelligence and power. They pay their tribute not just in subscription fees, but in the invaluable currency of their own secrets. Every query and document uploaded by their workforce represents a forfeiture of corporate privacy, allowing big tech to analyze and learn from the very information that constitutes their competitive advantage. In this new era, even billions in capital cannot guarantee you an audience with the king. The relationship is starkly reminiscent of Pope Clement VII refusing King Henry VIII’s appeal for a divorce; some requests, regardless of the petitioner’s power, will simply be denied. At the bottom of this pyramid are the end-users, the employees and customers who unknowingly feed this vast, hierarchical loop.

For corporate leaders, this is more than just a new technology to adopt; it is a profound political and infrastructural entanglement. The strategy of simple integration is a path to permanent dependency. Without sovereignty over their own computing infrastructure, the dream of model autonomy is just that, a fantasy. The boardrooms that once charted the course of global commerce must now chart a new one, navigating a world where they no longer rule. Their future independence may depend on radical new approaches: banding together in private alliances or investing in national-scale GPU initiatives. But one thing is clear: the era of unquestioned corporate power is over. They now kneel at a new throne, having traded their secrets for a seat at the table.

Read our full Report Disclaimer.