Nvidia’s power play: How Jensen Huang got Trump to rethink the China AI chip ban

Jul 17, 2025 - 17:46

0 0

Nvidia’s power play: How Jensen Huang got Trump to rethink the China AI chip ban

Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. You can sign up to receive this newsletter every week here.

Nvidia’s Huang says chip bans aren’t the way to deal with China

Nvidia founder and CEO Jensen Huang has been active on the government relations and lobbying front, and now he’s got something big to show for his efforts: the Trump Administration has agreed to lift a ban on selling Nvidia H20 AI chips to China.

Huang met with leaders in both Washington and Beijing, arguing that the AI revolution is a tide that will lift all boats—that AI technology can boost business productivity, raise the standard of living, and improve GDP for both the U.S. and China. He emphasized that the best way for America to maintain an edge in the AI race is to ensure the world’s AI models and apps run best on chips made by a U.S.-based company.

The U.S. (under Biden) initially began restricting sales of Nvidia’s most powerful chips to China in an effort to slow Beijing’s AI ambitions. The Trump Administration later doubled down, effectively banning sales of the H20 back in April. As a result, Nvidia reported a loss of about $2.5 billion in sales during its quarter ending in April, and projected it would miss out on another $8 billion in the quarter ending in July.

Huang apparently persuaded the Trump Administration to reverse course. His argument likely sounded something like this: “Our mission, properly expressed, is that in order for America to have AI leadership is to make sure that the American tech stack is available to markets all over the world so that amazing developers, including the ones in China, are able to build on the American tech stack so that AI runs best on the American tech stack,” Huang said during a recent interview with CNN’s Fareed Zakaria. Huang also noted that half of the world’s AI researchers “are in China and Chinese.”

Huang seems to be suggesting that the U.S. can retain technological dominance by controlling the platform AI runs on—similar to how it maintains financial dominance by ensuring most world trade is based on the dollar. There may be some truth to that. But it raises important questions: What does such control actually mean? Will the U.S. be able to dictate how the Chinese use the chips? No—no more than it did when DeepSeek used the H20 to build its world-class models. Is Huang implying that the U.S. could collect information about Chinese AI activities through these chips (as the U.S. once accused China of doing through Huawei)? That seems very doubtful.

There’s no doubt that Nvidia and its shareholders benefit when the world’s AI is built on its chips and software—but is America really better off? And if Huang truly believes the best AI chips shouldn’t be restricted, doesn’t it follow that he’ll also ask the Trump Administration for permission to sell China its most powerful Blackwell chips, too?

Fears grow that the U.S. government will use AI to surveille

Six months into a chaotic second Trump presidency, new reports have emerged suggesting the government is increasingly interested in using AI tools to track and profile U.S. residents. According to multiple whistleblowers and insiders, agents of Elon Musk’s DOGE are actively working to build a centralized, cross-agency database of Americans’ personal information—some of it highly sensitive.

The Washington Post reported in May that DOGE is rapidly constructing a centralized database that includes Social Security numbers, medical records, and tax files—doing so without regard for federal data privacy rules, and without standard oversight or even interagency agreements.

From the outset, DOGE has pushed past barriers and sidelined individuals to gain access to data stored at the Treasury, Office of Personnel Management, Social Security Administration, Health and Human Services, and the Departments of Education and Labor, reports the Brookings Institution

Meanwhile, concerns are also growing about how other agencies may be using AI to expand surveillance capabilities. ProPublica reported this week that the Internal Revenue Service is now developing a computer program that would give Immigration and Customs Enforcement (ICE) deportation officers unprecedented access to confidential tax data belonging to millions of American taxpayers, including their home addresses. In the past, ICE requested IRS data only for individuals it was actively investigating—typically no more than a dozen at a time. The new system could serve as a mass surveillance tool, possibly using AI, to identify new deportation targets. Due process may be a secondary concern.

Adding to the unease, FedScoop reported last week that the General Services Administration is considering using an AI model developed by Elon Musk’s xAI to process the personal data of American citizens.

Palantir (cofounded by Trump ally Peter Thiel) has become deeply embedded within agencies across the federal government. Its AI is used for data integration, analysis, and decision-making at defense and intelligence agencies, as well as FEMA, ICE, and HHS. Critics have raised concerns about the breadth and depth of data Palantir can access, and the lack of transparency regarding how its systems function.

After 9/11, Palantir began addressing the government’s urgent need to make sense of the vast volumes of intelligence data it was collecting on potential terrorist operatives and events both domestically and abroad. Since then, the use of Palantir’s platform has only grown—and it could easily be leveraged to form deep profiles on regular American citizens.

AWS launches a one-stop shop for enterprise AI Agents

Amazon’s AWS cloud division is placing a big bet on AI agents. At this week’s AWS Summit in New York City, the company unveiled “AI Agents and Tools,” a new section within the AWS Marketplace designed as a kind of concierge service for businesses looking to buy, deploy, and manage AI agents. The store will feature agents from AWS, as well as third parties like Anthropic, IBM, Perplexity, and Salesforce. Typically, AI agents can store large amounts of information about a company and its workflows, and can reason through tasks.

For existing AWS customers, the platform will likely simplify the process of integrating AI agents with AI models—allowing both to reside within the same secure cloud environment as their data. Amazon AWS is bundling everything companies need—databases, security tools, IT support, and deployment infrastructure—into one streamlined experience. Businesses will be able to describe their automation needs in plain English to an AI-powered search tool and receive customized recommendations on which agents are best suited for the job.

Gartner predicts that agents will automate half of all business decisions by 2027. And no company wants to fall behind while competitors gain new efficiencies. However, building custom agents from scratch can be a major challenge for corporate IT departments, often requiring significant additional infrastructure and integration work. The new AWS agent platform and marketplace could help eliminate those hurdles.

AWS is optimistic about the potential. “It upends the way software is built,” said AWS VP for Agentic AI Swami Sivasubramanian at the announcement. “It also introduces a host of new challenges to deploying and operating it, and potentially most impactfully, it changes how software interacts with the world—and how we interact with software.”