The United States and China are both moving to close the main route Chinese AI developers use to reach Nvidia chips: data centers in third countries. US senators have called for tighter limits on Nvidia's sales to China

1

2

, and Beijing is consulting on export controls that could bar Chinese companies from running their most advanced AI models in data centers outside China

1

2

.

The route matters because, per the report, Nvidia remains the dominant supplier of the chips Chinese AI and robotics firms need, and Chinese founders say so themselves

1

2

. Since those chips cannot be bought at home, developers get at least part of their access through data centers in third countries

1

2

.

The deal that collapsed

Tencent, the Chinese AI giant, reportedly pulled out of a $7 billion deal to lease Oracle data center capacity in Southeast Asia

1

2

. Both Washington and Beijing put pressure on the company to walk away

1

2

. Neither Tencent nor Oracle has confirmed the reported withdrawal

1

2

. The deal was a route to Nvidia chips that are unavailable inside China

1

2

.

The senator push follows the departure of former Nvidia CEO Jensen Huang, who stepped down after an affair with an alleged Chinese intelligence agent

1

2

.

The cheap-model bet

While the compute door narrows, Chinese AI firms are competing on price rather than frontier capability. The bet is that most customers want the cheapest model that does the job

1

2

. Per investor Larry Li, the price gap comes from lower labor, talent, hardware and energy costs in China, plus a large pool of engineering graduates and locally made hardware

1

2

.

Fintech company Revolut says it cut its compute costs by 90% after moving to a Chinese AI model

1

2

. That figure is the company's own claim; no independent audit has checked it

1

2

.

"They choose whichever model that does particularly a piece of work the best and at the best cost possible," said Fred Sun, Tencent's Europe general manager

1

2

.

Li, who has spent decades in Silicon Valley, says the squeeze may backfire on Washington. "I think that will speed up the process of isolation. But this really may not be bad for innovation. So that actually force other countries including China, to innovate on their own," he said

1

2

.

Robots as the other front

The physical side of the ecosystem is already at scale. Chinese factories installed more industrial robots over the past five years than every other country combined

1

2

. Embodied AI, the field of putting AI into machines that act in the real world, raised $13 billion in China in the first half of 2026

1

2

.

Source: Reuters

Source: Reuters

It is a harder problem than chatbot AI, because a robot has to interpret what it sees, what it hears and the physical space around it, not just text

1

2

. Xynova, a Chinese startup developing robotic hands, is one of the companies trying. "These hands can solve the labor shortage problem," said CEO Alex Xia, whose hands are aimed at repetitive work like packing and factory tasks

1

2

. Xia says his company watches the sanctions closely to make sure it obeys the rules in every jurisdiction it operates in

1

2

.

The report's conclusion is that the two AI ecosystems are developing in parallel, and users in third countries may end up having to pick a side

1

2

.

© 2026 TheOutpost.AI All rights reserved