At Nvidia's GTC conference this month, Jensen Huang was asked about Dario Amodei's warning that AI could eliminate half of all entry-level white-collar jobs within five years. His response, delivered to CNBC's Jim Cramer, was as confident as it was revealing.
"Because you're out of imagination," Huang said, addressing the hypothetical of a company that would use AI to cut headcount. "For companies with imagination, you will do more with more."
It is a compelling line. It is also, at this point, empirically contested by the behaviour of the largest technology companies on earth, companies that are, by any reasonable measure, not short on imagination.
Key Points
- Jensen Huang (Nvidia) argues that companies using AI to cut jobs simply lack imagination. Meta, Amazon, and Microsoft collectively eliminated over 46,000 roles while doubling AI investment.
- Dario Amodei (Anthropic) projects that AI could eliminate 50% of entry-level white-collar jobs within five years, with unemployment rising to 10-20%.
- The Dallas Federal Reserve found unemployment rising nearly three percentage points among workers aged 20-30 in AI-exposed occupations since early 2025.
- 98% of C-suite executives surveyed expect AI to cause headcount reductions in the next two years. These are plans, not forecasts.
- Huang profits from AI adoption regardless of its labour market impact. Amodei has proposed taxing his own company to mitigate it. The incentive structures differ.
What Amodei actually said
Dario Amodei, the CEO of Anthropic and one of the most credible voices on the capabilities and risks of frontier AI, did not offer a vague warning about disruption. He offered a specific one.
Within five years, Amodei estimates, AI could eliminate 50% of entry-level white-collar jobs. Unemployment could rise to between 10 and 20%. The vector of displacement he identifies is not limited to a single sector or skill set. It is AI's cognitive breadth, its ability to perform knowledge work across finance, consulting, law, and technology simultaneously. Workers displaced from one field will not simply retrain and move into another, because the next field is being automated in parallel.
Amodei has also proposed a structural response: a "token tax" requiring AI companies to contribute 3% of revenues to government redistribution programmes. He has been direct about his reasoning. "We, as the producers of this technology, have a duty and an obligation to be honest about what is coming."
That framing deserves attention. The CEO of one of the world's most advanced AI laboratories is stating, on the record, that the technology his company builds is likely to cause serious economic harm, and that the industry has a moral obligation to say so. Huang's response is to call that assessment a failure of imagination.
The companies doing "more with less"
The difficulty with Huang's argument is that the evidence does not point in his direction. It points in Amodei's.
In January 2026, Amazon eliminated 16,000 corporate roles, with its own leadership explicitly citing AI and automation as drivers of the decision. This is not a company that failed to invest in AI. Amazon Web Services is among the most significant AI infrastructure providers in the world. Amazon's AI budget is not constrained. Its headcount is.
Meta announced it would cut approximately 15,000 employees, roughly 20% of its corporate workforce, while simultaneously doubling its AI capital expenditure budget to $135 billion for 2026. The logic is not difficult to read: the AI investment is replacing the labour. That is not a failure of imagination. That is the business model.
Microsoft cut more than 15,000 positions during 2025 while committing $80 billion to AI infrastructure. These are not companies that lack ambition or vision. They are companies executing a straightforward substitution: AI capacity in, human headcount out.
The numbers
Meta, Amazon, and Microsoft have collectively eliminated over 46,000 positions since 2025 while collectively increasing AI infrastructure investment by hundreds of billions of dollars. In each case, AI was cited by leadership as a contributing factor. Huang's framework does not explain this pattern. It explains it away.
The data from the labour market
The evidence is not confined to corporate announcements. The Dallas Federal Reserve published findings in early 2026 showing that workers aged 20 to 30 in AI-exposed occupations had seen unemployment rise almost three percentage points since early 2025. That is a material shift in a short timeframe, concentrated precisely in the cohort that Amodei identified as most vulnerable: young workers entering knowledge-work careers.
A separate survey of C-suite executives found that 98% expect AI to cause headcount reductions over the next two years. These are not pessimists. These are the people making the decisions. Their expectations are, in effect, plans.
Geoffrey Hinton, who won the Nobel Prize in Physics in 2024 for his foundational work on neural networks, has been consistent in his assessment: AI will increase unemployment while driving higher corporate profits. He is not speaking from ignorance of the technology. He helped build it.
The problem with optimism from the inside
Jensen Huang is not a neutral observer. Nvidia manufactures the hardware that powers the AI revolution. Every GPU sold to a company replacing workers with AI systems is revenue for Nvidia. Every data centre built to run large language models at scale runs on Nvidia silicon. Huang's optimism about AI's impact on employment is, structurally, also an argument for his company's continued growth.
That does not make him wrong. But it is a relevant fact when weighing his confidence against Amodei's alarm. One of these men profits from the adoption of AI regardless of its labour market consequences. The other has proposed taxing his own company's revenues to mitigate those consequences. The incentive structures are not equivalent.
Huang's vision of a future where companies employ "hundreds of thousands of AI agents alongside humans" is coherent. It is also a vision that describes a future in which the ratio of AI agents to human workers is the central economic variable, and in which that ratio is moving in one direction. MIT and Stanford research into what emergent AI agent civilizations actually look like when left to run suggests the trajectory is not hypothetical. The game industry has already lived through the early chapters of this story. So have the young programmers whose hiring rates have collapsed even as AI investment in software development has surged.
"We, as the producers of this technology, have a duty and an obligation to be honest about what is coming."
Dario Amodei, CEO of Anthropic
What is actually at stake
The debate between Huang and Amodei is not really a disagreement about technology. It is a disagreement about who bears the cost of the transition.
Huang's "do more with more" scenario is genuinely possible. There are historical precedents for new technologies creating more work than they eliminated. The internet and smartphones generated entire categories of employment that did not previously exist. Huang is not inventing his position from nothing.
But Amodei's concern is precisely that this transition is different in structure, not just in scale. Previous automation was narrow: it displaced specific physical tasks in specific industries, and workers could retrain into adjacent roles or adjacent sectors. AI's cognitive breadth means the displacement is simultaneous and sector-spanning. There is no adjacent lane to move into when the adjacent lane is also being automated.
The workers who are most exposed to this are not abstract economic units. They are people currently in entry-level white-collar roles, or students preparing to enter them. They are the 74% of game design students who told GDC 2026 researchers they are worried about their job prospects. They are the young workers in AI-exposed occupations whose unemployment is already rising faster than the aggregate. They are the 16,000 people who left Amazon with a severance package and a world that now has fewer roles for them than it did a year ago.
Jensen Huang says they work for companies that lack imagination. Dario Amodei says the companies that employ them are making rational economic decisions, and that those decisions will accelerate. The evidence, so far, supports Amodei.