The Trump administration struck deals with Nvidia and AMD last August that were supposed to channel a slice of their China chip sales directly into government coffers. Nearly a year later, the US government has collected exactly zero dollars from either company.

The arrangements required both chipmakers to pay 15% of revenues from sales of specific AI chips to China as a condition for receiving export licenses. The administration also floated the idea of bumping the take to 25% of profits on advanced AI chip exports, layering a profit share on top of the revenue share.

Nvidia reported roughly $50 million in H20 chip sales to China under these licensed agreements by late 2025. And even on that modest sum, the government’s 15% cut, roughly $7.5 million, never actually arrived. As of mid-2026, no payments have been reported from either Nvidia or AMD.

No rules, no enforcement

One reason the payments haven’t materialized is strikingly simple. No formal regulations implementing these revenue-sharing arrangements were ever published as of late 2025. The deals were announced, confirmed publicly by President Trump, and then left in a kind of regulatory limbo. Without published rules specifying payment timelines, penalties for non-compliance, or even basic accounting procedures, the revenue-sharing model looks more like a press release than a policy.

The administration has been moving away from the direct-subsidy model that defined the Biden-era CHIPS and Science Act of 2022, which poured federal money into domestic semiconductor manufacturing. Instead, the current approach seeks to extract value from chip companies rather than invest in them. The most dramatic example: discussions around acquiring a 10% equity stake in Intel, reportedly valued at approximately $10 billion.

What this means for investors

For anyone holding Nvidia or AMD stock, the immediate financial impact of these arrangements is negligible. A 15% revenue share on $50 million in sales isn’t moving the needle for companies with market capitalizations in the hundreds of billions.

The real risk is regulatory uncertainty. If the administration decides to formalize these agreements with binding regulations, the terms could become more demanding. The jump from 15% of revenue to 25% of profits was already floated, and there’s nothing stopping further escalation. Companies that depend on government export licenses are, by definition, at the mercy of whatever conditions get attached to those licenses.

The Intel equity stake discussion adds another layer of complexity. If the government starts taking ownership positions in semiconductor companies, it changes the relationship between regulator and regulated in ways that are difficult to predict. A government shareholder has different incentives than a government regulator, and the potential for conflicts of interest is substantial.

For the broader semiconductor sector, the key variable to watch is whether formal regulations ever materialize. Published rules would signal that the administration is serious about this model and intends to enforce it. Continued ambiguity suggests the arrangements may quietly fade, replaced by whatever trade policy framework comes next.