Nvidia's CEO Just Used His First-Ever Tweet to Take a Side. It's Good News for Your EA Career.

2026-07-315 min read

Last reviewed: July 24, 2026. Based on Jensen Huang's July 24 X post, the "Open Weights and American AI Leadership" letter, and signatory analysis.

Jensen Huang joined X on June 26, 2026. He has 200 billion dollars and runs the most valuable company on Earth. He waited a month before posting anything.

When he finally posted, it wasn't a product announcement. It wasn't a keynote clip. It was an open letter signed by 25 companies telling Washington not to kill open-weight AI models.

The letter is titled "Open Weights and American AI Leadership." It is published on nvidia.com. It is also on microsoft.com. The signatories include Meta, Palantir, Andreessen Horowitz, Hugging Face, IBM, Y Combinator, and Mistral. The letter argues that open-weight models — AI that anyone can download, inspect, and run on their own hardware — are essential infrastructure, like open-source software was in the 1980s.

Here is who did not sign: OpenAI. Anthropic.

The AI industry just split in public. The CEO of Nvidia chose his first words on social media to take a side. That is not a random act of corporate communications. That is a structural market signal from the person with the most to gain or lose from how this fight plays out.

The split

On one side: the open-weights coalition. Nvidia, Meta, Microsoft, and two dozen others. Their argument is straightforward. Open models spread AI capability across the economy instead of concentrating it in a few hands. They create competition across chips, cloud, applications, and services. They let customers own their own value instead of renting it from a single provider. They make the whole stack more secure by letting more people inspect it.

On the other side: OpenAI and Anthropic. The two companies that have raised the most money on the premise that they will build the best models and charge for access. Their business model depends on being better than what you can download for free. Every open model that matches their capability is a direct threat to their revenue.

This split was always coming. The surprise is how fast it arrived. Jensen Huang posting for the first time, taking a side, and publishing the coalition letter on Nvidia's own domain — this is the moment the split became legible to everyone.

Why Nvidia wants open models

Nvidia's interest is the easiest to read. Open models mean more people running AI on more hardware. Every downloaded model is a potential GPU customer. Every fine-tuned variant is another training run. Every local deployment is another inference workload.

But the letter's arguments go beyond Nvidia's commercial interest. It explicitly defends distillation — the technique of using one model's outputs to train another — as "a widely used technique for model improvement, evaluation, and validation." It warns Washington not to "conflate legitimate model-development techniques with misappropriation."

This matters because distillation is exactly how Chinese AI labs caught up to American frontier models. The narrative from some corners has been that Chinese labs "stole" American AI through distillation. The open-weights coalition is saying: no, that's just how the technology works. The letter is a preemptive defense of the technique against regulation that would ban it.

What this means for tax professionals

If you are studying for the EA exam, or considering it, the industry split is good news. Here is why.

When an industry fractures into competing camps, the value shifts away from the platforms and toward the people who can operate regardless of which platform wins. You do not need to know whether OpenAI or Meta builds the better model. You need to know tax law. You need to be able to represent clients before the IRS. You need a credential that the government, not a venture capitalist, issued.

The EA credential is platform-agnostic. It does not matter whether the AI that helps you prepare a return was trained by Nvidia hardware or Google TPUs. It does not matter whether the model is open-weight or closed. What matters is that your name is on the return and your credential is on the line.

When technology platforms fight each other, the people who own the liability layer win. The platforms need each other dead. You just need clients who need returns filed.

The letter's timing

The letter dropped on July 24, 2026. Washington is currently considering new AI export controls and model restrictions. Anthropic's S-1 is expected any week now. OpenAI's is expected this fall. The open-weights coalition published their argument at the exact moment the closed-model labs are about to show their financials to the public.

If Anthropic's S-1 reveals sustainable unit economics, the closed-model narrative strengthens. If it reveals the reverse-economics documented by industry analysts — losing money on every API call, burning cash faster than revenue grows — the open-weights argument becomes stronger. The letter is positioned for both outcomes.

The letter also lands just as Chinese open-source models cross 60 percent of OpenRouter token share. The letter does not mention China explicitly. It does not need to. The argument that "open weights create competition" is also the argument that Chinese open-source models are legitimate competitors, not thieves.

The bottom line

The AI industry is splitting into two camps. The CEO of the most valuable company in the world just used his first-ever social media post to pick a side. The fight is about whether AI capability concentrates in a few closed providers or diffuses across an open ecosystem.

Neither outcome threatens the EA credential. The credential is a federal license. It grants representation rights. It carries legal liability. These are not features of a technology platform. They are functions of law.

When the platforms fight, the credential holders watch from above.

Dojo Challenge · strike before you scroll

Retrieve it before the explanation retrieves it for you.

Bernice prepared her client's 2025 tax return on January 30, 2026. The client, Rudolf, had claimed three children as his dependents, claiming EITC based on the three children, and also claimed head of household status. Bernice didn't ask any questions about the children, or collect any documents. Rudolf's return was later audited by the IRS, and it was discovered that he had claimed his girlfriend's children, who were not related to him (Rudolf and his girlfriend are not married). All of the EITC credits for the three children, and the HOH status on Rudolf's return were disallowed. What kind of penalty can apply to Bernice, the preparer of the return in this case?

Enter a full practice round →Check readiness

Keep building this concept · 3 related guides

Chinese AI Just Killed the Silicon Valley Monopoly Playbook — And That's Good for Your EA CareerAmerican AI companies planned to burn cash, kill competitors, then jack up prices. Chinese open-source models now account for the majority of AI tokens used by American firms. Local models running on laptops match frontier performance. The monopoly endgame is off the table. Credentials become the new scarce resource.Becoming an Enrolled Agent From India: The 2026 Path, Costs, and Real Career OutlookIndia produces more EA candidates than any country outside the US. Remote testing is now available. Over 1,500 US tax outsourcing firms operate in India. Here's the real path, what it costs, and what the job market actually looks like.The AI Bubble Is Deflating. Here's Why That's Good for Tax Careers.While tech stocks crash and AI layoffs surge past 45,000, the Enrolled Agent path is quietly absorbing the career changers that the hype cycle created. The bubble bursting doesn't hurt tax prep. It helps.

Find your weak topics →Open the 50-unit path →