The most important AI regulatory story this week isn’t a court ruling or a new bill. It’s a meeting.

President Trump is bringing Nvidia’s Jensen Huang, Apple’s Tim Cook, and Elon Musk to a summit with Xi Jinping, according to reporting from Ars Technica. The gathering may force the administration to reconsider its posture on semiconductor tariffs and Taiwan. And if it does, the downstream effects on US AI policy, export control law, and the global chip supply chain would be substantial.

Jensen Huang being in the room is not incidental. It’s the whole point.

What the Export Control Regime Actually Does

To understand why this matters, you need a quick primer on how the US currently controls AI chips.

The Bureau of Industry and Security, a division of the Commerce Department, administers the Export Administration Regulations (EAR), found at 15 CFR Parts 730-774. The EAR governs what American companies can sell abroad, and to whom. When BIS adds a company to its Entity List, US firms are effectively barred from selling it controlled goods without a license that, in practice, won’t be granted.

The Biden administration used this framework aggressively starting in October 2022, restricting the export of advanced semiconductors and chipmaking equipment to China. The rules were expanded again in late 2023 and 2024, with BIS targeting not just chips above certain performance thresholds but also the manufacturing equipment used to make them. The key tool for extending US reach beyond American borders is the Foreign Direct Product Rule, which lets BIS control foreign-made goods if they were produced using American technology or software. That rule is why a chip made by a Taiwanese factory on Dutch equipment still falls under US jurisdiction if American IP is embedded in the process.

The Trump administration inherited this framework. It hasn’t dismantled it, and in some respects has pushed it further. But tariffs, which operate under a different legal authority, have created a complicated parallel pressure. Tariffs are blunt instruments. Export controls are surgical. The administration has been running both simultaneously, and the friction between them is now playing out at the highest diplomatic level.

Nvidia’s Position Is Genuinely Complicated

Jensen Huang is not a passive participant here. Nvidia occupies a near-monopoly position in the AI training chip market. The H100 and its successors are the workhorses of large-scale model training, and China represents a massive potential market. BIS has required Nvidia to produce downgraded versions of its chips for the Chinese market, variants like the A800 and H800, which were specifically engineered to comply with export thresholds. When BIS tightened the rules again, those chips also got restricted.

Every time the rules tighten, Nvidia loses revenue. Every time they loosen, Nvidia gains it. There’s nothing nefarious about Huang being at this summit. He’s the CEO of the company most directly affected by the policies under discussion. But that also means his presence is a signal. The administration is at least entertaining the idea of giving ground.

The national security argument for the controls is straightforward: advanced AI chips accelerate military AI applications, and the US government doesn’t want China’s People’s Liberation Army training frontier models on American silicon. The Commerce Department, the Defense Department, and the intelligence community have all supported the restrictions on those grounds. Congress passed the CHIPS and Science Act in 2022 partly to shore up domestic semiconductor manufacturing against exactly this kind of geopolitical vulnerability.

The counterargument, which Nvidia and others have made, is that the restrictions are leaky. Chinese firms have found workarounds, including routing purchases through third countries. American companies lose the business, Chinese developers find other means, and the US competitive position in chipmaking doesn’t actually improve. It’s a real argument. It just doesn’t mean the controls should disappear entirely.

Taiwan Is the Third Rail

The Ars Technica report flags Taiwan as the other variable the summit might touch. This is where things get genuinely precarious from a regulatory standpoint.

Taiwan Semiconductor Manufacturing Company, TSMC, fabricates the overwhelming majority of the world’s most advanced chips, including Nvidia’s. Apple designs its own silicon and TSMC makes that too. If anything disrupts TSMC’s operations, whether through military conflict, a change in US-Taiwan relations, or diplomatic pressure, the entire AI hardware supply chain breaks. There is no American or European alternative at TSMC’s process nodes. Not yet, and not for years.

The CHIPS Act included significant funding aimed at building out domestic fab capacity, and TSMC has committed to building fabs in Arizona. But those facilities are not a near-term substitute for Taiwan’s production capacity. When Tim Cook is in that room, the Taiwan question isn’t abstract. It’s about whether the phones and laptops Apple sells will continue to be manufacturable at the scale Apple requires.

Any US concessions on Taiwan policy, even rhetorical ones, would raise serious concerns in Congress and at the Defense Department. The Taiwan Relations Act, passed in 1979, commits the US to providing Taiwan with defensive arms and treating threats to Taiwan as a matter of “grave concern.” It doesn’t require military intervention, but it establishes a framework of support that multiple administrations have maintained. Signaling flexibility on that framework to extract trade concessions would be a significant departure.

What Happens Next

The meeting hasn’t happened yet, and no deal has been struck. “Agreement in principle” is the language of the telecom story this week, not this one. But the composition of the US delegation tells you what’s on the table.

If the administration does pivot on semiconductor restrictions, expect Congressional pushback fast. The CHIPS Act has bipartisan support. So does a hawkish posture toward China on technology. Senators on both sides have been willing to go further than the executive on chip controls, not less far. Any rollback would likely trigger hearings and possibly legislation trying to lock the controls in place.

The other thing to watch is BIS rule-making. Export control changes that flow through the regulatory process take time and create administrative records that can be challenged. Tariff changes can happen faster, by executive action. If the administration wants to signal goodwill to Beijing without formally unwinding BIS restrictions, adjusting tariffs is an easier lever to pull. That distinction matters for how durable any shift would be, and for who can challenge it.

The AI chip supply chain is the most important piece of physical infrastructure in the current AI development cycle. The legal and regulatory frameworks built around it over the last four years represent a real policy choice about who should have access to the most powerful AI hardware. That choice is now being renegotiated, partly behind closed doors, by people who have a financial stake in the outcome.

That’s not a scandal. But it is worth watching very carefully.