global geopolitics

How the fight over AI and China cost Jensen Huang his company, and why the official explanation does not survive scrutiny

Opinion Analysis | November 2006

The story begins with a fact that most people would find strange if they stopped to think about it, because a man who built one of the world’s most valuable companies was removed from it in less than a week. Jensen Huang founded NVIDIA thirty-three years ago and turned a small maker of graphics cards into the company whose chips now power almost every serious artificial intelligence system on the planet. On 8 November 2026 the board declared complete confidence in him, and four days later the same board removed him and did not even leave him a seat at the table. The official reason given to the public was a spy investigation, and the investigation was leaked to the press before any charge had been laid or any evidence tested. A genuine security investigation into a figure of this importance would have been kept secret until it was complete, because publicising it alerts everyone involved and destroys the chance of finding out what actually happened.

The mechanism of removal deserves attention because it reveals how the company was structured to be vulnerable. NVIDIA has a single class of shares and no special voting rights for its founders, and its board has eleven directors, so six votes were enough to remove the chief executive without giving a reason. Companies like Meta protect their founders with shares that carry extra votes, which means Mark Zuckerberg could not be removed by his board even if the board wanted to act. When a founder gives up those protections, the arrangement is praised as good governance in quiet times and becomes a fatal weakness when powerful outside interests decide that the founder must go. The absence of a protective structure was not the cause of the removal, but it was the condition that made the removal possible within a single week.

The reason outside interests wanted him gone is not difficult to establish once the business of the company is understood. NVIDIA makes the hardware that trains and runs artificial intelligence, and its chips and its software environment are used by both American and Chinese developers, which means the company sits at the exact point where the two powers collide. American policymakers wanted to go further than they had already gone, by restricting where the chips can be used, by cutting Chinese developers off from the software needed to run them, and by directing the company’s output toward military programmes. Huang resisted those measures because he ran a commercial business that sold the best possible hardware to everyone who wanted to buy it, and because the restrictions would have damaged the company he had spent three decades building. The campaign against him began when his commercial judgement came into conflict with the strategic objectives of the state, and the spy investigation supplied the justification for removing him.

The spy investigation itself follows a pattern that intelligence services have used for a very long time, in which a threat is announced publicly in order to shape opinion rather than to uncover anything. If the American government genuinely believed that the head of its most important chip company had been compromised by China, it would have moved quietly to contain the damage and to identify everyone connected to the operation. Announcing the investigation guaranteed that anyone involved would destroy evidence and cut contact with everyone else, which means the announcement cannot have been intended to produce a real intelligence result. Western news organisations work closely with intelligence agencies and rarely publish sensitive security material without approval, so the appearance of the story across multiple outlets at the same moment indicates coordination rather than independent reporting. The most likely purpose of the leak was to create public pressure that would make the board’s decision easier and would discourage Huang from resisting it.

The second analytical lens concerns the chokepoint, which is a point in a supply chain that everyone must pass through and that a single actor can therefore control. NVIDIA’s chips and its software environment constitute such a chokepoint for artificial intelligence, because advanced development is very difficult without access to them and because switching to alternatives takes years and enormous investment. The United States has been using that chokepoint as an instrument of pressure against China, and the instrument works best when the company that controls it cooperates fully with the strategy. A chief executive who lobbied against restrictions and who found ways to keep selling to Chinese customers reduced the pressure that the chokepoint was supposed to apply. Removing him and replacing him with an interim finance chief, who has neither the standing nor the inclination to lobby against the policy, restored the chokepoint’s effectiveness at a stroke.

The third lens concerns structural power, which means control over the underlying systems that everyone else has to use rather than control over any single transaction. The security system, the production system, the financial system and the knowledge system are the four structures through which power is exercised in the modern world, and NVIDIA occupies a position in all four at once. Its chips are treated as a security asset, its manufacturing relationships determine what can be produced and where, its valuation shapes the capital markets, and its software environment determines what developers can build. A company that holds all four positions simultaneously is far too important to be left to the judgement of a single founder, which is exactly why the state moved when that founder’s judgement diverged from its own. The state did not need to build a rival company, because it needed only to change the person at the top of the existing one.

The fourth lens concerns political warfare, which describes the use of methods short of direct military force to achieve strategic goals against an adversary. The operation against Huang combined several of those methods at once, including a fabricated security scare, a summons to appear before Congress, coordinated media coverage and a social media campaign that relocated blame onto a foreign country. Political warfare is effective because each instrument reinforces the others, and because the target is not only the individual but also the public that might otherwise object to what is being done. The congressional summons converted a commercial dispute into a legal obligation, the media coverage converted a rumour into apparent fact, and the social media campaign converted a corporate decision into evidence of foreign manipulation. Together they produced a result that no single instrument could have produced alone.

The fifth lens concerns securitisation, which describes the process by which an ordinary policy question is reclassified as an existential threat that requires emergency measures. Artificial intelligence has been reclassified in exactly this way, and once a technology is treated as essential to national survival, the ordinary rules of commerce no longer apply to it. Export controls that would once have been unthinkable became routine, restrictions on foreign access became a security necessity, and a chief executive who questioned those measures became a security risk rather than a businessman with a different view. The spy investigation was the moment at which the securitisation became personal, because it converted a disagreement about policy into an allegation about loyalty, and allegations about loyalty cannot be answered with commercial arguments.

The sixth lens concerns elite circulation, which describes how powerful individuals rise and fall while the structure of power itself remains unchanged. Huang was replaced, but the position he occupied and the system within which that position exists were not altered in any way, and the new occupant of the position will operate under the same constraints that he did. The comparison with Intel is instructive here, because Intel removed its chief executive in 2018 and replaced him with a finance officer, and the company then missed the entire mobile and artificial intelligence transition. NVIDIA is more exposed than Intel was, because its business depends on a technology roadmap that requires deep technical judgement at the very top. A finance officer facing Wall Street questions about the next generation of chips is not in a strong position to defend a roadmap that a technologist spent years constructing.

The seventh lens concerns hegemonic competition, which describes the struggle of a dominant power to preserve its position against a rising challenger. American policymakers believe that artificial intelligence is the technology that will decide whether the United States remains the leading power in the world, and they have organised a great deal of policy around that belief. The removal of Huang should be understood as one move within that larger contest, and the direction of the move is unmistakably toward confrontation rather than accommodation. Anyone who claims that Washington is seeking a settlement with Russia or China should account for the fact that the same week produced the removal of the one executive who had been lobbying hardest against the China restrictions. The window for selling specially designed chips to China is now likely to close, and that outcome was the point of the exercise rather than an unintended consequence.

The social media campaign that blamed Israel for the takeover deserves separate treatment because it illustrates how the blame-shifting works in practice. The campaign rested on a single sentence from a 2023 interview in which NVIDIA’s finance chief said that her working day begins with meetings involving the company’s Israeli executives. The full article explains that NVIDIA bought an Israeli company called Mellanox for 6.9 billion dollars in 2019, that roughly 3,200 of its employees are now based in Israel, and that the morning call is simply a matter of time zones. When it is nine in the morning in California it is seven in the evening in Israel, so the finance chief must speak to her Israeli subordinates before they go home for the night. The call consists of an American executive instructing employees of a company that NVIDIA owns, which is the exact opposite of the claim that was being circulated.

The second element of the campaign rested on an Israeli newspaper article listing three Israeli technology figures who might one day qualify to run a company like NVIDIA. The article itself states that there is no indication any of them has been offered the job or wants it, and two of the three do not even work at the company. A local newspaper speculating about local talent is not evidence of a foreign takeover, and treating it as such tells us more about the beliefs of the people sharing it than about anything that happened. The claim that a country of fewer than ten million people, dependent on American support for its existence, could quietly seize control of the most important technology company in the United States does not survive even a moment of ordinary scrutiny.

The reason this matters is that the blame-shifting is not an accident but a documented strategy, and the document in question has been public since 2009. The Brookings Institution published a study of American options toward Iran with a chapter titled “Leave It to Bibi,” which recommended encouraging an Israeli strike in order to deflect international criticism and Iranian retaliation away from the United States. The logic of that recommendation has since been applied far beyond Iran, and it now extends to any American action that would look bad if the public understood who had ordered it. When a corporation is stripped of its founder, and the public is told that a foreign country did it, the American role disappears from the story and the real decision-makers escape scrutiny. The people sharing that story believe they are resisting power, and in fact they are protecting it.

The corporate documents tell a quieter version of the same story, and they are worth reading because they show what the company itself expected. NVIDIA’s own press release from 2019 describes the Mellanox purchase as an acquisition by NVIDIA of a company it would then own and control, which is the ordinary language of corporate expansion. The finance chief’s interview describes a manager managing, not a foreign power infiltrating, and the succession article describes journalists speculating, not a board conspiring. The gap between what those documents say and what was claimed about them is the space in which the blame-shifting operated, and closing that gap requires only that people read the original rather than the summary. That is a modest requirement, and it is remarkable how rarely it is met.

The costs of all this fall on people who had no part in the decision and no say in it. The workers of the semiconductor supply chain in Taiwan, South Korea and elsewhere produce the hardware under conditions that are often hazardous, and their communities would be the first to suffer if the competition over chips turned into something worse. The Chinese developers and students excluded from advanced computing lose access to tools that their counterparts elsewhere take for granted, which entrenches a division that the postwar order was supposed to close. The American public pays for the programme through taxes and absorbs the risk of a confrontation that it never voted for, and it is told a story about foreign manipulation instead of a story about who actually decided. The people of the Global South, who are rarely mentioned in any of this, face a new version of an old division in which the technology of the age is owned by a small number of countries and rented to everyone else.

The question of who benefits has a clear answer once the ownership of the relevant institutions is examined. The defence contractors that supply the military programmes, the cloud providers that host the artificial intelligence systems, the chip companies that benefit from restricted competition and the financial institutions whose valuations depend on continued technological dominance all gain from the direction of travel. The think tanks that produce the policy papers are funded by those same corporations, and the officials who move between the government and the industry carry the assumptions of both into every room they enter. This is not a conspiracy in the sense of a secret meeting, because the coordination happens openly through shared interest and shared personnel. It is a class interest expressed through institutions, and the removal of a dissenting executive is one of its routine operations.

The most likely path forward is continued tightening, because the logic that produced the removal of Huang points toward further restrictions rather than toward accommodation. The interim leadership will avoid controversy, the next permanent chief executive will be selected partly for reliability on China policy, and Congress will attach further conditions to export licences in the defence authorisation legislation. China will accelerate its effort to build an independent technology stack, which will take years and will consume enormous resources, and the result will be a more fragmented global computing economy with higher costs for everyone. The risk of the coming period is not that either side wins quickly but that both sides spend a decade and a great deal of money building parallel systems that could have been shared. The people who will pay for that duplication are the ordinary users and workers who have no seat at any of the tables where it is being decided.

The deeper question raised by this episode concerns who has the right to decide how a society’s most important technology is used. A founder who built a company and resisted its conversion into an instrument of state policy was removed within a week, and the public was told a story that had nothing to do with the real reason. If decisions of that magnitude can be taken without public deliberation and then explained away with a fabricated security scare, then the claim that the public governs itself is difficult to sustain. The remedy is not a single policy change but a habit of attention, in which people read the original documents, ask who benefits, and refuse the explanations that relocate blame onto whoever is least able to have done the thing. That habit is not sufficient on its own, and it is the ground on which anything else has to be built.

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

Brookings Institution (2009) Which Path to Persia? Options for a New American Strategy toward Iran. Washington, DC: Brookings Institution Press.

Dunlap, C. J. (2001) ‘Law and military interventions: Preserving humanitarian values in 21st century conflicts’, Carr Center for Human Rights Policy Working Paper. Cambridge, MA: Harvard Kennedy School.

Farrell, H. and Newman, A. L. (2019) ‘Weaponized interdependence: How global economic networks shape state coercion’, International Security, 44(1), pp. 42–79.

Galtung, J. (1971) ‘A structural theory of imperialism’, Journal of Peace Research, 8(2), pp. 81–117.

Gilpin, R. (1981) War and Change in World Politics. Cambridge: Cambridge University Press.

Globes (2023) ‘Trillion dollar woman’, Globes. Available at: https://en.globes.co.il/ (Accessed: 15 November 2026).

Intel Corporation (2018) Intel CEO Brian Krzanich resigns, 21 June. Santa Clara: Intel.

Kennan, G. F. (1948) ‘The inauguration of organized political warfare’, Policy Planning Staff Memorandum, 4 May. Washington, DC: US Department of State.

Mearsheimer, J. J. (2001) The Tragedy of Great Power Politics. New York: W. W. Norton.

Mills, C. W. (1956) The Power Elite. New York: Oxford University Press.

NVIDIA Corporation (2019) NVIDIA to acquire Mellanox for $6.9 billion, 11 March. Santa Clara: NVIDIA.

Pareto, V. (1916) The Mind and Society. Florence: Barbera (English edition: New York: Harcourt, Brace, 1935).

Strange, S. (1988) States and Markets. London: Pinter.

Ynet (2026) ‘After Jensen Huang: The Israelis who could inherit Nvidia’s AI empire’, Ynet, 12 November. Available at: https://www.ynetnews.com/ (Accessed: 15 November 2026).