World news 126K views since yesterday

China can rent Nvidia computing power without importing the chips

Andrew Chang's analysis examines overseas cloud access, physical smuggling and why export restrictions can still limit China's AI capacity.

Channel: CBC News ·

CanadaStory of November 14, 2026Trari Newsroom

Why this story? One of CBC News's most watched videos since yesterday, with 125,562 views.

CBC News: The news division of the CBC, Canada's public broadcaster. 4.8M subscribers and 3.4B views.

How China keeps getting its hands on Nvidia's AI chips | About That

What did you think?

Things we can learn from the video

Press to jump

1:03Chang explains Nvidia's Blackwell Ultra architecture and why its chips are prized for AI development.
2:49U.S. export restrictions target advanced AI chips and chip-manufacturing equipment shipped to China.
4:25Chang recounts alleged physical chip-smuggling schemes pursued by the U.S. Department of Justice.
6:30Experts argue that restrictions still constrain China's computing scale, despite access through black markets.
7:08Chinese firms can rent overseas Nvidia computing capacity; at 09:20, Chang examines legislation aimed at restricting remote access.
10:30Scrutiny surrounding Megaspeed frames the difficulty of establishing where restricted hardware ultimately goes.

Keeping an advanced chip out of China does not necessarily stop a Chinese company from using it. Renting access to computers abroad offers another route, one that Andrew Chang identifies as a central gap in Washington's effort to restrict access to Nvidia's AI technology.

Canada's CBC News, the Canadian Broadcasting Corporation's news service, examines that distinction in "How China keeps getting its hands on Nvidia's AI chips | About That". The report moves beyond the scrutiny surrounding Jensen Huang's departure to ask how restrictions work when computing power can be delivered across borders without the hardware moving at all.

Chang begins with why the chips matter. He explains Nvidia's Blackwell Ultra architecture, while Chris Miller, author of Chip War, describes the company's hardware as the gold standard for training artificial intelligence models. U.S. restrictions cover both advanced AI chips and equipment used to manufacture them.

The report then separates two ways around those barriers. One involves physically moving restricted hardware: Chang recounts alleged smuggling schemes pursued by the U.S. Department of Justice. TECHnalysis Research analyst Bob O'Donnell explains that restrictions on sought-after technology create demand for a black market.

The other route does not require importing a chip. Chang explains that Chinese firms can remotely rent computing capacity from facilities outside China that run Nvidia hardware. He examines U.S. legislative efforts, including the Remote Access Security Act, intended to restrict that access.

But access is not the same as unlimited capacity. Miller argues that Google alone has more computing power available than China as a whole. Erich Grunewald, a senior researcher at the Institute for AI Policy and Strategy, makes the distinction explicit: "What matters here is not just the quality of the chip, but also the quantity."

The final section considers scrutiny surrounding Megaspeed and Huang's proximity to President Donald Trump. O'Donnell identifies the practical enforcement problem beneath the political questions: establishing where products ultimately end up can be difficult.

Published on Trari

Most watched today ›