“Nobody needs atomic bombs, everybody needs AI” was the arguement that got him removed within four days, and the machinery that replaced him is already metering the water, the power and the thought itself
Opunion Analysis | November 2022
The sentence that Jensen Huang spoke on the BG2 podcast in September 2025 was simple and it was dangerous because it described a world in which the most important technology of the age would be a common utility rather than a private estate. “Nobody needs atomic bombs,” he said. “Everybody needs AI.” The contrast was deliberate, because nuclear weapons are the technology of concentrated destruction and artificial intelligence is, or could be, the technology of distributed capability (PodLexicon, 2025). He went further in the same conversation, describing AI as an infrastructure company’s product and arguing that every nation should build its own sovereign capability so that its language, its history and its people were represented in the systems they would come to depend upon (PodLexicon, 2025). He did not say that every nation should buy its intelligence from a handful of American firms, and he did not say that access should be metered by subscription, and he did not say that the water and the electricity required to run the machines should be taken from communities that had never consented. He said everybody needs it, which is the language of a right rather than the language of a market.
The removal came with the speed that institutions reserve for emergencies they have themselves created. The board held an emergency meeting late into the night and announced that Huang would step down immediately from his roles as chief executive and board member, and the official statement praised his contributions without disclosing the reasons for his departure (KBS, 2026). The timing is the part that history will find difficult to explain, because the resignation followed four days after he admitted an extramarital affair and in the wake of allegations involving Chinese espionage that had been circulating without charge or adjudication (KBS, 2026). A board that had declared complete confidence in its founder months earlier removed him in the time it takes to hold a single meeting, and the interim appointment of a chief financial officer signalled that the priority was continuity of the balance sheet rather than continuity of the mission (KBS, 2026). The man who said everybody needs AI was replaced by a caretaker who will not say it, and that replacement is the quiet act of enclosure that the rest of this essay will describe.
The material basis of the enclosure is a concentration of compute that no single firm has held since the height of the industrial trusts. NVIDIA controlled approximately sixty-three per cent of global AI compute capacity in the second quarter of 2026 and sold sixty-six billion dollars worth of AI chips in that quarter alone, and it held more than eighty per cent of AI chip sales in a market where its software ecosystem constitutes a lock-in that no competitor has replicated at scale (NDTV, 2026). The company is not merely a supplier; it has been described as the central bank of the AI economy, the node through which financing, allocation and access all pass (Semi Analysis, 2026). A firm that controls the compute layer controls the conditions under which every other firm, every university, every hospital and every government must operate, and a founder who wanted to distribute that access was a obstacle to the rentier logic that the concentration makes possible. The board did not remove him because the chips stopped working, and it removed him because the distribution he advocated would have reduced the scarcity on which the pricing depends.
The resource dimension of the enclosure is the part that touches people who will never use a chatbot. A single hundred-megawatt hyperscale facility can consume as much water as eighty thousand people, often in water-stressed regions, and campaigners have warned that allowing a handful of firms to monopolise energy and water supplies at the expense of communities is as reckless as it is foolish (Forbes, 2026). Global electricity consumption by data centres rose about seventeen per cent in 2025 and consumption by AI data centres grew by about fifty per cent, and the International Energy Agency estimates that total data-centre electricity use could reach nine hundred and fifty terawatt-hours by 2030 (ECPS, 2026). These are not abstract figures; they are the water that does not reach a farm, the electricity that does not reach a hospital and the grid capacity that is diverted from households to machines that are owned by firms whose shareholders are not the people who bear the cost. The promise of abundance is being financed by an enclosure of the commons, and the enclosure is proceeding without a vote and without a treaty and without the consent of anyone who will pay for it.
The rentier model that follows from this concentration is already visible in the business models of the firms that sit on top of the compute. A ChatGPT subscription is a commodity like any other, and the companies selling AI subscriptions have been marked up excessively while the costs of large language models increase exponentially and subscription revenues increase only linearly (Varoufakis, 2026). The unsustainable arithmetic of selling access as a commodity has pushed the industry toward what Yanis Varoufakis calls the technofeudal sector, in which cloud capital extracts rents from every interaction and in which the interface knows you better than you know yourself and can price your dependence accordingly (Varoufakis, 2026). The AI assistant that remembers everything you have ever said and done and wished for is not a public servant, and it is a landlord of your own cognitive life, and the rent is charged in the currency of attention, data and subscription fees. A founder who wanted to make the technology a utility was standing in the way of a business model that depends on keeping it a toll road.
The language of the Great Reset and Agenda 2030 enters here not as a conspiracy theory but as a description of a governance architecture that is being built in plain sight. Critics have documented how digital control systems, biometric identities, smart cities and programmable currencies are converging to produce a future in which freedom becomes a subscription and dissent is quietly deleted (Chamseddine, 2026). Others describe a digital feudal matrix in which signing up to the AI assistants as a service means the death of decision fatigue for compliant citizens, and in which the monthly stipend arrives credited to an AI prompt account while the rent is due (Gentleman-Hobbs, 2026). These are not the fantasies of people who fear technology; they are descriptions of the political economy that the technology is being used to build, and they are consistent with the direction of travel that the compute concentration, the water extraction and the subscription model all point toward. The AI that Huang described as a universal need is being converted into a universal dependency, and the conversion is being managed by the same institutions that removed him.
The human cost of the enclosure is borne by the people who were supposed to benefit from the promise. A smallholder farmer in Zimbabwe who could use an AI system to read the weather and identify a crop disease cannot afford the subscription, and the data centre that trains the model consumes water from a catchment that his community also depends upon (Forbes, 2026). A student in a public school in Lagos or Jakarta or La Paz who could use an AI tutor cannot access the compute, because the compute is priced for the institutions that can pay and the model does not speak her language well enough to teach her (PodLexicon, 2025). A patient in a rural clinic whose diagnosis could be assisted by an AI system waits for a connectivity that the same firms have not found profitable to provide. The promise was that AI would be the great equaliser, and the reality is that it is becoming the great divider, and the divider is being built by the firms that control the chokepoint and financed by the resources that the chokepoint consumes.
The removal of Huang is therefore not a corporate footnote, and it is the moment at which the choice between a commons and a cage became visible. He said everybody needs AI, and he was removed in four days by a board that did not disclose its reasons and by a state apparatus that had been circulating espionage allegations without charge (KBS, 2026). The interim chief executive will not repeat the argument, and the next permanent chief executive will be selected partly for reliability on the questions that matter to the state, and the company will continue to sell the compute that the world needs at a price that the world cannot refuse. The enclosure will proceed, and the water will be extracted, and the subscriptions will renew automatically, and the language of the Great Reset will describe it all as a gift. The only thing that can interrupt this trajectory is a politics that treats intelligence as a common inheritance rather than a private estate, and that politics begins with the recognition that a founder who said everybody needs AI was removed because he meant it.
Authored By: Global GeoPolitics
Thank you for visiting. This is a reader-supported publication. If you believe journalism should serve the public, not the powerful, and you’re in a position to help, becoming a PAID SUBSCRIBER truly makes a difference. Alternatively you can support by way of a cup of coffee:
https://buymeacoffee.com/ggtv |
https://ko-fi.com/globalgeopolitics |
Bitcoin: 3NiK8BoRZnkwJSHZSekuXKFizGPopkE7
References
Chamseddine, G. (2026) The 2030 Enigma: Behind the Curtain of the Global Agenda. London: Feltrinelli.
ECPS (2026) ‘The algorithmic green grab: digital sovereignty, populist promises and the ecological costs of AI’, European Center for Populism Studies, 3 November.
Forbes (2026) ‘Campaigners join forces to raise environmental concerns over data centers’, Forbes, 27 February.
Gentleman-Hobbs, N. (2026) ‘The new social contract is the digital feudal matrix’, LinkedIn, 8 January.
KBS (2026) ‘Jensen Huang ousted at Nvidia’, KBS News, 14 November.
NDTV (2026) ‘NVIDIA: The company that rivals nations’, NDTV, 11 November.
PodLexicon (2025) ‘NVIDIA: OpenAI, future of compute, and the American dream – BG2 w/ Bill Gurley and Brad Gerstner’, PodLexicon, 26 September.
Semi Analysis (2026) ‘Nvidia as the central bank of AI’, *Semi Analysis*, 13 November.
Varoufakis, Y. (2026) ‘Can we escape AI slavery? Only the left can tame big tech’, Yanis Varoufakis, 12 February.

Leave a Reply