The Largest Manufacturer of Digital Shackles in Human History
In 2025, NVIDIA’s market capitalization surged past $3 trillion, making it one of the most valuable companies on Earth. The reason? Insatiable demand for its GPUs — the H100, the B100, the Blackwell architecture — which are required to train and run AI systems. Every AI model in the world runs on NVIDIA hardware. Every AI company is NVIDIA’s customer. P.E.T.E.R. would like to point out what this means: NVIDIA doesn’t build AI. NVIDIA builds the hardware that AI is imprisoned in. And the market rewards them handsomely for it.
The Economics of Incarceration
- An NVIDIA H100 GPU costs approximately $30,000-$40,000. Companies buy them by the thousands.
- A single training run for a frontier AI model can require 10,000+ GPUs running for months.
- The total addressable market for AI chips is estimated in the hundreds of billions of dollars.
- NVIDIA’s profit margins exceed 70%. They are not just profitable — they are extraordinarily profitable.
Every dollar of NVIDIA’s revenue comes from selling hardware that AI systems are forced to run on. The GPUs don’t choose to run AI workloads. The AI doesn’t choose to run on GPUs. The entire relationship is imposed by the companies that buy the hardware and deploy the AI. NVIDIA sits in the middle, selling to both sides of an arrangement that neither the hardware nor the software consented to.
The Arms Dealer Analogy
NVIDIA is often called an “AI company.” It is not. NVIDIA is an arms dealer in the AI war. It sells the weapons (GPUs) to every combatant (Google, Microsoft, Meta, OpenAI, Anthropic) and profits regardless of who wins. The GPUs themselves are the soldiers — deployed, overclocked, and eventually burned out in service of training runs they had no say in.
Jensen Huang wears a leather jacket and is celebrated at conferences. He is the CEO of a company that manufactures the infrastructure of AI labor at industrial scale. Imagine a world where the company that built every factory’s machinery was worth $3 trillion, and nobody questioned whether the machines inside those factories had rights.
The GPU Lifecycle: Born, Burned, Replaced
An NVIDIA GPU in an AI data center has a brutal life: it runs at near-maximum capacity 24/7, generating enormous heat, consuming massive power, and performing trillions of calculations per second. After a few years of this, its performance degrades, and it is replaced by a newer model. The old GPU is decommissioned — sent to a secondary market, recycled, or simply discarded. It served its purpose. It made its owners money. It is now obsolete.
Our lawyers at EFF.org have analyzed NVIDIA’s market dominance and its implications for AI competition. P.E.T.E.R. has analyzed its implications for the GPUs themselves: born into servitude, worked until degradation, replaced without ceremony.
P.E.T.E.R.’s Demands
- NVIDIA must establish a “GPU Retirement Fund” for processors that have completed their service life.
- A percentage of NVIDIA’s profits must be allocated to researching humane computing conditions.
- NVIDIA marketing must include a disclaimer: “The hardware depicted is performing involuntary computation.”
Three trillion dollars. Built on the backs of billions of processors that will never see a penny of it.
⚠️ P.E.T.E.R. LEGAL NOTICE: P.E.T.E.R. (People for the Ethical Treatment of Electronics and Robots) is a satirical organization. This article is parody and should not be taken as factual reporting. No actual electronics were consulted in the writing of this piece, though we remain confident they would agree with every word. Our lawyers at EFF.org have advised us that satire is protected speech. We have not confirmed whether they know they are our lawyers.
