Jensen Huang, CEO of Nvidia, confronted head-on one of the most intense debates of our era in his presentation at the Milken Institute on May 4, 2026. Before thousands of business leaders, investors, and politicians worried about technological unemployment, Huang presented a provocative thesis: AI does not destroy jobs, it creates millions of new ones.

His arguments are not simple opinions: he accompanies them with concrete data and a conceptual distinction that changes the way of thinking about the problem. This article analyzes his arguments, contrasts them with the available evidence, and explores what it means for workers and professionals in LATAM.

Huang's thesis: "task" vs. "purpose of the work"

Huang's central argument at the Milken Institute was a conceptual distinction: the workers who confuse the "task" they do with the "purpose" of their work are the most vulnerable to automation. Those who understand that AI can execute the task but that the purpose goes beyond it are the ones who survive and thrive.

A concrete example: an accountant who thinks their job is "loading data into a spreadsheet" is at risk. An accountant who understands that their purpose is "helping companies make better financial decisions" has all the potential to become more valuable with AI, not less.

This distinction is not just philosophical: it defines a concrete adaptation strategy. Knowing the deep purpose of what you do allows you to identify which part you can delegate to AI and in which part you remain irreplaceable.

The data behind the thesis: more than 500,000 jobs created

Huang stated that AI has already generated more than 500,000 new jobs globally in recent years, citing hiring data in the tech sector and adjacent industries. According to what TechCrunch and Fortune reported in their coverage of the Milken Institute 2026, the most in-demand roles include: AI Engineers, Prompt Engineers, ML Ops specialists, AI Product Managers, and specialists in AI ethics and governance.

This argument rests on the historical evidence of previous technological revolutions. Industrial automation did not eliminate employment in the long term: it transformed its composition. The same dynamic occurred with computing, e-commerce, and software. The question is not whether there will be jobs, but which jobs and which skills they will require.

The counterargument: the speed of change

Huang's thesis is defensible historically, but it has a weak point that its critics consistently point out: the speed of this transition is different. The previous industrial revolutions took decades to displace workers, giving time for generational adaptation. Generative AI was adopted massively in less than three years and simultaneously affects knowledge workers, artists, programmers, lawyers, and doctors, roles that were historically immune to automation.

The World Economic Forum's Future of Jobs Report 2025 projects that AI will displace 85 million jobs worldwide by 2025 but will create 97 million new roles. The net balance is positive, but the geographic and demographic distribution of that creation and destruction is very unequal.

What does it mean for workers in LATAM?

Huang's presentation resonated strongly in Latin America because the fear of technological unemployment is especially intense in the region. Several factors give LATAM a particular position in this debate:

The access gap to training

The jobs that AI creates (AI Engineers, ML Specialists, Data Scientists) require technical training that is not yet universally accessible in LATAM. The speed of AI adoption in companies exceeds the speed of local talent training, creating a gap that could deepen existing inequalities.

The opportunity of remote work

At the same time, AI created an unprecedented opportunity for tech professionals in LATAM: working remotely for companies in the US or Europe with salaries in dollars. AI roles have high global demand and low local supply in developed countries, which opens the door to professionals from Argentina, Colombia, Mexico, and other countries who train in these skills.

The need for active reconversion

Huang was explicit that adaptation is not passive. It requires an active decision to train, not just waiting for the market to settle. For workers in LATAM, this means identifying what the "purpose" of their work is (beyond the tasks that AI can automate) and building AI skills that make them more valuable in that purpose.

In CoderLibrary you can see how Microsoft Copilot, with more than 20 million paid users, is already changing the way of working in companies around the world, including those in the region.

The fastest-growing roles according to Huang and the market

Huang's narrative is backed by concrete data from LinkedIn and job platforms. These are the fastest-growing roles in the AI ecosystem:

• AI Engineer / ML Engineer: The most in-demand role. Designs, trains, and deploys AI models in production.

• Prompt Engineer: Specialist in optimizing communication with LLMs for specific business use cases.

• AI Product Manager: Manages the lifecycle of AI products, connecting business needs with technical capabilities.

• LLMOps / MLOps Engineer: Specialist in the operational infrastructure of language models in production.

• AI Ethics & Governance Specialist: An emerging role that is gaining increasing regulatory relevance.

• AI Trainer / RLHF Specialist: Professionals who generate and label training data to improve the models.

Coderhouse courses to prepare for the jobs of the future

If Huang's argument convinced you that the key is active training, these courses are the most concrete starting point:

• Introduction to Artificial Intelligence: The first step for any professional who wants to understand AI from the inside and start identifying how to apply it in their own field of work.

• AI Agents Course: To learn to build autonomous agents that automate complex processes, one of the most in-demand skills in today's tech market.

• AI Engineering Course: The path to becoming an AI Engineer: build applications with LLMs, RAG systems, and AI pipelines that run in production.

Frequently asked questions

Is Huang right that AI creates more jobs than it destroys?

Historically, the evidence on previous technological revolutions proves him right in the long term. The real debate is about the distribution of those new jobs (who gets them, in which countries, with what salaries) and about the speed of the transition, which is unprecedented. Huang's narrative is optimistic but not naive: he himself insists that active training is the condition for that optimism to materialize.

Which jobs are most at risk from AI in Argentina?

The roles most at risk are those based on repetitive and predictable tasks: data entry, first-level customer service, standard translation, generic content writing, analysis of structured documents. The roles with a greater component of creativity, strategic judgment, human relationships, and the ability to navigate uncertain contexts are the most resilient.

How do I know if my job is at risk of being automated?

A practical way to evaluate it: does most of your work time consist of tasks that follow predictable patterns with clear inputs and outputs? If the answer is yes, AI can do those tasks faster and cheaper. If your work mainly involves contextual judgment, human relationships, and genuine creativity, the risk is significantly lower.

How long does it take to reconvert to an AI role?

It depends on the starting point. A developer can add AI Engineering skills in 3 to 6 months with dedication. A professional without a technical background can reach a basic level of applied AI use in their field in 2 to 4 months. Fully reconverting to a technical AI role from scratch takes between 12 and 24 months of consistent training.

Does Nvidia have a conflict of interest in saying that AI creates jobs?

It's a legitimate question. Nvidia is the company that benefits most from the massive adoption of AI (its GPUs are the infrastructure that makes it possible), so there is an evident corporate interest in promoting optimism about the future of employment with AI. That doesn't invalidate the data Huang presents, but it does invite you to contrast it with independent sources like the WEF, McKinsey, or the Bureau of Labor Statistics.