Nvidia Tightens Cloud Partner Rules on China While Oracle Build-Out Raises Funding Questions

Nvidia has told cloud partners around the world to reduce their reliance on Chinese customers, requiring that at least 51% of services under future contracts go to companies outside China, according to a Financial Times report. Reuters carried the report today but could not independently verify the details.

The directive, which the FT says has been communicated verbally over the past several months and applies globally, covers new contracts only. Existing computing-capacity rental agreements remain untouched, meaning the permissible share of Chinese customers would top out at 49%. The distinction matters: the report describes no immediate withdrawal from all business with Chinese clients — rather, Nvidia would narrow the future customer mix of its partners. Nvidia's sales and regional teams are expected to enforce the rules.

Cloud providers in Singapore warned in the report that finding enough non-Chinese customers could prove difficult, and that strict enforcement might hurt their business. For Nvidia, the exposure therefore extends to partners outside China whose demand depends on Chinese users.

Export Controls Reach Down the Customer Chain

The reported requirements tie into stricter US restrictions on advanced AI technology and a crackdown on smuggling networks. The Commerce Department issued corresponding guidance in May, covering products that include Blackwell chips. Nvidia had already cut its list of approved Asian customers by more than 50% in July. Its vetting includes data-center visits, contract reviews and conversations with end users.

The controls thus reach beyond the immediate delivery of hardware to how cloud providers' customers use the computing power — potentially widening the commercial impact of export rules past individual chip sales. Existing rental contracts would be protected under the report, while future deals face tighter limits. Concrete revenue effects cannot yet be quantified.

Oracle Supercluster Plans and the Financing Question

The China rules land alongside Nvidia's expanding cloud footprint, which brings its own questions about how AI infrastructure gets paid for. On October 27, Nvidia and Oracle announced they would widen their collaboration on large computing clusters, scaling Oracle Cloud Infrastructure superclusters built on NVIDIA Vera Rubin NVL72 to several gigawatts. The two companies also made those systems and Vera processors generally available on the OCI platform.

Availability and expansion plans should be kept apart. General availability has already been announced; the build-out to the stated scale remains a project. The partnership describes a concrete use case for Nvidia's technology but does not answer the separate question of how large infrastructure projects will be financed.

That is where Melius Research weighed in with an assessment published yesterday. The firm kept its buy rating but lowered its price target to $315 from $420, citing valuation pressure because AI infrastructure financing will depend more heavily on backstops through 2028. The combination is what matters for investors: Melius did not withdraw its recommendation, only cut the target, so a positive view of the stock and greater caution on its valuation sit side by side.

Credit Backstops and Antitrust Scrutiny

Reuters-reported talks over multi-billion-dollar credit guarantees for Nscale fit the same financing theme. Nvidia could thereby support funding for a data center in West Virginia intended for Anthropic. According to the report, the support was not finalized and terms could change — such talks are not a financing commitment, and Nvidia's potential contribution should not be treated as a decided obligation.

Separately, Reuters reported on October 28 that the US Department of Justice had sent inquiries about the planned $12.93 billion acquisition of Hugging Face, examining possible effects on competition in AI. The requests are part of a broader review of Nvidia's acquisitions, investments and sales practices.

Nvidia has scheduled its earnings call for the third quarter of fiscal 2027 on November 17, 2026. Until then, cloud expansion, financing questions and competition oversight form the framework for valuing the stock. The Oracle announcement stands for planned additional infrastructure; neither the discussed credit guarantees nor the planned acquisition should be judged as completed deals.