Nvidia’s CEO, Jensen Huang, has sharply criticized the U.S. policy of restricting advanced chip exports to China, arguing that the move backfired by accelerating Chinese technological advancement.

Speaking at the Computex tech conference in Taipei, Huang described the Biden-era export controls as counterproductive and welcomed President Donald Trump’s decision to scrap some of those curbs.

“Incredible things have happened in the last year as a result of the export controls,” Huang told reporters. “The local companies are very, very talented and very determined, and the export control gave them the spirit, the energy, and the government support to accelerate their development.”

Huang pointed to China’s rapid progress in artificial intelligence and chip manufacturing as unintended consequences of Washington’s clampdown on AI chip exports.

“I think, all in all, the export control was a failure,” he said.

The U.S. had sought to curb Beijing’s access to high-end semiconductor technology amid rising geopolitical tensions, citing concerns over the potential military applications of artificial intelligence. But the move cost American firms, including Nvidia, a significant share of their most lucrative overseas market.

Nvidia, once dominant in China with an estimated 95% market share for high-end AI chips, has seen that figure cut nearly in half. We went from basically having 95% share to having 50% share, Huang said.

He added that while the original intent of the policy was to hinder China’s AI capabilities, it instead fueled a surge in domestic innovation. It’s labeled the unintended consequence of a policy that didn’t work.

Now, Trump appears poised to reverse that course. His administration is reportedly planning to replace the current tiered export control regime with a global licensing system that would allow technology trade under government-to-government agreements. The approach is seen as a more strategic and flexible way to maintain U.S. influence in global tech markets while avoiding unintended backlash.

Huang believes President Trump realizes it’s exactly the wrong goal to try to isolate U.S. technology from the world because it undermines our leadership. This is because if U.S. technology is not available to the rest of the world, they will build it themselves.

Indeed, that has already happened. Chinese firms, with strong state baking, have responded to the curbs by making rapid strides in tech self-sufficiency. Huawei’s breakthroughs in chipmaking and the rollout of its own HarmonyOS mobile operating system—designed to replace Google’s Android—have been cited as clear examples of how restrictions pushed China to innovate faster.

The Biden administration’s AI export controls, imposed in 2022 and 2023, particularly targeted Nvidia’s top-performing chips, such as the A100, H100, and more recently, the H800 and A800 variants tailored for Chinese clients. In response, Nvidia began designing downgraded versions to remain compliant. The company is now preparing to release a variant of its powerful Blackwell GPU architecture with reduced memory capacity to meet export standards.

But even with these workarounds, the Chinese market has grown more competitive. Huang acknowledged that domestic chipmakers such as Huawei and startup firms, with Beijing’s support, have filled the vacuum.

At Computex, Huang stressed the importance of maintaining access to global markets. He emphasized that the need for U.S. technology to be widely used.

China, for its part, has consistently condemned the U.S. curbs as discriminatory and politically motivated. The country’s commerce ministry reiterated this stance earlier this year, warning that it would “take resolute measures” to safeguard its interests if Washington continued to impose technology barriers.

As global AI competition intensifies, U.S. policymakers are facing an increasing challenge of how to constrain strategic rivals without undermining American firms or driving them into isolation.

Trump’s strategy signals a potential rebalancing, away from blanket restrictions and toward a model that preserves commercial advantage while allowing room for geopolitical negotiation. However, it is not clear whether this course correction will recover lost ground or simply mark a shift in the broader technological power balancing.

Tesla’s long-anticipated robotaxi service is finally making its debut, starting in Austin this June, in what CEO Elon Musk says will be a cautious but fast-paced rollout that could see up to 1,000 autonomous Teslas operating in the city within a few months.

The move marks a significant milestone in Tesla’s vision to reshape urban mobility and carve out a new revenue frontier beyond electric vehicle (EV) sales.

“We’ll start with probably 10 for a week, then increase it to 20, 30, 40,” Musk said in an interview with CNBC on Tuesday. “It will probably be at 1,000 within a few months.”

This will be Tesla’s first real-world deployment of a robotaxi service, part of a broader push by Musk to revolutionize transportation. He has long pitched the idea of a fleet of fully autonomous Teslas that could earn money for their owners, operating like a self-driving Uber, whenever the car is not in personal use. For Tesla owners, this promises a potentially lucrative side hustle. For Tesla, it introduces a new business model that could generate recurring revenue while amplifying the brand’s dominance in mobility tech.

The launch will begin under strict controls. The pilot will be invite-only, and the robotaxis will operate on public roads but be geo-fenced to limited, safer regions of Austin.

“We will geo-fence it,” Musk said. “It’s not going to take intersections unless we are highly confident it’s going to do well with that intersection. Or it will just take a route around that intersection.”

Teleoperators — remote human supervisors — will be on standby to intervene when the autonomous system gets stuck or faces an edge-case scenario. While competitors like Waymo and Zoox also use similar systems, Tesla has historically taken a different approach to autonomy, relying on camera-based systems and neural networks rather than expensive lidar and detailed 3D maps.

Musk has promised that Tesla will scale rapidly once the initial fleet proves itself. After Austin, the company plans to roll out robotaxis to other cities, including San Francisco. By the end of 2026, Musk claims there could be more than 1 million self-driving Teslas on the road in the United States alone — a bolder rehash of a promise he first made in 2019. At the time, Musk claimed there would be 1 million Tesla robotaxis by the end of that year. That goal was missed, and Musk later conceded that punctuality was not his “strong suit.”

Regulatory Hurdles Ahead

But scaling won’t be simple. Tesla’s push to launch robotaxis across multiple states will run headlong into a patchwork of regulations. California, for instance, home to Musk and Tesla’s earliest Full Self-Driving (FSD) experiments, still hasn’t granted full approval to Tesla for robotaxi operations. Alphabet’s Waymo, on the other hand, has already made significant progress in cities like San Francisco, Phoenix, and Los Angeles, and has permission to offer fully driverless rides to paying customers.

“The approval process is very haphazard and sort of state-by-state, and sometimes city-by-city,” Musk said, voicing frustration at the regulatory fragmentation. “We really need to get national regulations in place.”

That lack of federal cohesion could delay Tesla’s rollout in key markets, giving competitors more time to entrench themselves. Waymo, for instance, already operates in dense urban environments and has quietly expanded its fleet while earning regulators’ trust. Unlike Tesla, Waymo has focused on building a purpose-designed self-driving vehicle from the ground up, equipped with lidar and operating under a highly conservative safety model. These choices may slow down development, but have helped avoid high-profile incidents.

Tesla’s Full Self-Driving software remains under public and regulatory scrutiny. A recent test by Business Insider comparing Waymo and Tesla’s supervised FSD showed a Tesla running a red light at a complex intersection in San Francisco. Musk dismissed the test as flawed, saying it “made no sense,” but acknowledged Tesla’s robotaxis will avoid high-risk intersections until system confidence improves.

The robotaxi pilot, while limited in geographic scope, will serve as a crucial test of Tesla’s AI-driven approach to autonomy. Unlike Waymo’s laser-scanned maps and tightly controlled routes, Tesla is betting on fleet-collected real-world data and neural network learning to scale rapidly and cheaply. If it works, Tesla could leapfrog the more cautious, hardware-heavy approaches used by rivals.

A New Business Model

Tesla’s robotaxi venture isn’t just about advancing self-driving technology — it’s also about redefining Tesla as more than an automaker. Musk envisions a future where most Tesla owners can opt in to the autonomous network, allowing their vehicles to work as driverless taxis while they’re at home or at work. This model would transform every Tesla into a potential revenue-generating asset, something no other car company currently offers.

The company has said that eventually, these robotaxis could function without any manual controls, built specifically for autonomy. A new robotaxi design is expected to be unveiled at Tesla’s upcoming event on August 8.