A new draft regulation circulating in Washington would fundamentally alter how American companies sell advanced AI chips abroad. The proposal, a stricter evolution of a Biden-era framework, mandates that firms like Nvidia and AMD obtain government permits for most international sales of high-performance accelerators. This replaces previous tiered country lists with a case-by-case review, aiming to prevent chips from reaching Chinese data centers through third-party diversions.

The policy, expected to take effect after a Trump-ordered delay expires this September, places immense logistical pressure on a global industry. Nvidia, which reported $35.1 billion in data center revenue for Q4 fiscal 2025, has long argued that restrictive controls fuel demand for Chinese alternatives like Huawei's Ascend chips. Yet U.S. officials point to documented cases of chips being rerouted via shell companies in Malaysia and the UAE as justification for tighter oversight.

For cloud providers and nations building sovereign AI, the permit system's speed will be everything. Projects with tight construction schedules, like Microsoft's $80 billion global infrastructure push, cannot absorb months-long licensing delays. Countries from Saudi Arabia to France have lobbied for exemptions, fearing their national AI initiatives will stall.

The Semiconductor Industry Association has labeled the approach "unworkable," warning it cedes market share without strengthening security. The core tension remains: while controls may slow China's progress, they also accelerate its push for self-sufficiency. For now, companies are advised to bolster compliance teams and engage directly with regulators during the comment period. The era of unfettered AI chip trade is over, replaced by a new, uncertain regime of government scrutiny.

Source: Webpronews