Here is a sentence that would have read as science fiction ten years ago and is now simply a news item: American negotiators used the promise of access to Nvidia’s artificial-intelligence chips to help secure a preliminary peace deal between Armenia and Azerbaijan. The specific inducement — expanded approvals for chip purchases tied to an Armenian data-center project — had not been reported before. Officials have a name for the practice. They call it chip diplomacy.
The conflict it helped address is not a trivial one. Armenia, a mostly Christian country, and majority-Muslim Azerbaijan have clashed intermittently for more than 35 years, since before the collapse of the Soviet Union of which both were member republics. By some estimates nearly 40,000 people have died in the violence between them. When the current administration took office in 2025 it made peace in the South Caucasus a priority. Part of what closed it was permission to buy hardware.
This is not the first time, but it is a new kind
The administration has used AI hardware to negotiate agreements with the United Arab Emirates and Saudi Arabia as well. What makes the Armenia case notable, according to diplomats and experts quoted in the reporting, is that this is among the first known instances of using chips to help clinch an actual peace deal rather than a commercial or strategic arrangement. Along the way it deepened the relationship between the White House and the world’s largest chip designer, whose chief executive speaks with the administration regularly.
There is a reasonable case that this is good news — a genuine conflict moved toward resolution, and the currency was export licenses rather than weapons. There is also a straightforward observation to make about it, which has nothing to do with whether the policy is wise.
What happens when a product becomes a policy instrument
A company whose product is used as a bargaining chip in foreign policy has acquired a second business it did not apply for. Its addressable market is now partly set in Washington rather than by customers. Approvals can expand, as they did here, and they can contract. That variable will not appear on any income statement, will not be forecast by any analyst with confidence, and does not move with the ordinary drivers of the business.
This is the same lesson as the safety probe elsewhere in this edition, arriving from the opposite direction: some of the most consequential inputs to a large company’s future are decisions made by people who do not work there and do not answer to its shareholders. The correct response is not to avoid such companies — that would rule out most of the modern economy — but to hold them in sizes that reflect the presence of a variable you cannot model.
For a household portfolio the practical version is short. If a single company has become a large share of what you own — and after the last three years, for many people it has — then part of your retirement now depends on export policy. That is worth knowing deliberately rather than discovering later. It takes about fifteen minutes with the statements to find out how much of your plan is riding on decisions made in a building you will never enter.