Nvidia has advised some cloud customers to take on less business from Chinese companies. The guidance is aimed at encouraging providers, including Nvidia-backed neoclouds, to diversify who they serve.
The request is voluntary and applies to future contracts, not existing deals. Nvidia is also asking cloud customers to disclose the identities of the end users accessing its chips.
A request focused on future business
The Information reported the guidance, citing three people familiar with the matter. The Financial Times reported that the instruction was communicated verbally through Nvidia sales representatives and regional teams.
Those details point to a customer relationship approach rather than a new mandatory restriction. Providers are being asked to consider the composition of future business, while contracts already in place are outside the stated scope.
Asking for end-user identities adds another layer of visibility. A cloud provider may host computing capacity while a separate company uses it remotely, so knowing who is using the chips can help clarify where the demand comes from.
Why cloud access matters to export controls
U.S. export controls restrict chip shipments, but generally do not cover remote access through cloud services. That creates a gap between rules governing the movement of hardware and the ability to use computing power from a distance.
Chinese AI developers can access advanced Nvidia chips by renting compute capacity at data centers outside China. In that arrangement, the hardware remains at a foreign facility and the customer uses its capabilities remotely, without importing the chips themselves.
This distinction helps explain Nvidia's focus on cloud customers and end users. The issue described in the reports is not only where chips are physically shipped, but also which companies can draw on them through rented data center capacity.
Attention on access through foreign data centers
U.S. authorities had been looking into Chinese AI companies accessing chips through foreign data centers after allegations surfaced involving Moonshot AI, the developer behind the Kimi chatbot.
The source does not say that the allegations resulted in a finding of wrongdoing, or that the guidance is a government directive. It describes a voluntary request from Nvidia to certain cloud customers, against the backdrop of scrutiny over remote access.
For cloud providers, the request means that future commercial decisions may involve weighing customer diversity alongside demand for Nvidia-powered computing. Identifying the end users could also make it easier to understand who benefits from the capacity being offered, even when the chips stay abroad.
What the guidance could mean for providers
The guidance leaves providers with discretion: it is not mandatory, and existing agreements are not covered. Its practical effect will depend on how cloud customers respond when considering new contracts with Chinese companies.
The broader question raised by the situation is how export controls apply when access to advanced chips is delivered as a service. Restrictions on shipping hardware do not automatically settle the separate question of remote use, and the reports describe Nvidia seeking more visibility and a more varied customer base in that environment.