Export controls decide who can buy the most advanced AI chips and chipmaking tools, and in 2026 they're one of the biggest forces shaping the AI industry. The US restricts sales of top AI accelerators and equipment to China. Rules have shifted to allow some sales, such as Nvidia's H200, while China discourages its companies from buying US chips and pours money into domestic alternatives.
The current shape of things
- US controls limit exports of leading-edge AI chips and the equipment to make them, with licences and exceptions that change over time.
- Some loosening: the US decision to allow exports of Nvidia's H200 to China was a notable shift, though analysts at Oxford Economics don't expect it to slow China's localisation drive.
- China's response: reportedly discouraging purchases of Nvidia's China-specific chips, setting domestic sourcing goals, and weighing large incentive packages for local chipmakers.
- Everyone else is building more local capacity, including India, with fabs and packaging plants in Gujarat and Assam.
Export rules change frequently, so check the latest official guidance before making business decisions that depend on them.
Why a developer should care
- AI prices and availability depend on chip supply. Restrictions and stockpiling affect GPU cloud prices and capacity.
- Model choices are becoming regional. Some models and services are unavailable in some countries, and some are trained on different hardware stacks.
- Data and compliance rules follow geopolitics. Where your data and model run can matter for government and enterprise clients.
Building with less risk
Keep AI features portable: model calls behind one interface, a second provider ready, and data stored where your clients' rules require. That turns a policy change into a config change, not a crisis.
At NextGen Fusion, our AI and software development company in Lucknow, we design AI products with that portability from day one.
