A look at Sugon, a Chinese company at the heart of the country's surveillance efforts, and how it uses Intel and Nvidia chips to analyze video feeds at scale
Intel and Nvidia chips power a supercomputing center that tracks people in a place where government suppresses minorities …
New York Times
Context & Ripple Effects
This 2020 investigation documented the concrete end-use of American silicon: a supercomputing center built on Intel and Nvidia chips that analyzes video feeds to track people where the government suppresses minorities. It landed years before Washington's export-control regime matured, when such deployments were still largely unpoliced.
Intel and Nvidia are directly implicated as suppliers whose chips power a named surveillance operator, Sugon, giving regulators and human-rights investigators a specific supply chain to target rather than an abstraction.
Sugon's operations depend on continued access to US accelerators, making its tracking capability hostage to any tightening of export licenses or enforcement.
Second-order effects
Enforcement burden shifts onto the chipmakers themselves: Nvidia's global compliance build-out and its denials around chip diversion via data centers are direct responses to the end-use exposure stories like this one created.
Chinese institutions exposed as dependent on Nvidia and AMD parts face substitution pressure, pushing procurement toward domestic alternatives wherever performance allows.
Third-order effects
If the pattern holds, US export policy evolves from list-based restrictions toward end-use verification — with chipmakers absorbing the cost of policing their own customers, and gray-market channels persisting wherever verification lags.
The deeper structural shift is compute becoming a lever of statecraft: surveillance capability, AI progress, and chip access get treated as one strategic question by both Washington and Beijing.
The trend: American chips remain the compute backbone of China's surveillance infrastructure even as export controls tighten, forcing enforcement to migrate from sales bans toward end-use policing.
So what can this machine do? According to Sugon, the contractor that made it, as of 2018 it could connect to 10k video feeds an analyze 1,000 simultaneously. A gov't post said it could query 100 mln photos in a second. That made it the best image/video analysis machine in China. …
“I'm afraid in a few years time, #China companies + government will find their own way to develop chips and these capabilities,” @wang_maya said. “Then there will be no way to get a handle on trying to stop these abuses.” @hrw @hrw_chinese https://www.nytimes.com/...
The computer is repeatedly linked to “predictive policing.” In practice in Xinjiang that has meant sweeping people up based on their religious activities, travel history, and personal tech choices, like whether a person owns two phones or no phone. https://www.nytimes.com/...
China's big data AI police surveillance can “search 100 million photos in a second ... its computers [can] connect to 10,000 video feeds and analyze 1,000 simultaneously, using artificial intelligence.” Wake up, people. https://www.nytimes.com/...
Chips made by American companies are powering a supercomputing center in Xinjiang that's used to track people and process surveillance footage, reported by @paulmozur & Don Clark. https://www.nytimes.com/...
from my experience, @Intel is more open & forthcoming about preventing human rights-abusing applications of its tech, esp. re: China, however that is not the case with @nvidia
Top-end Nvidia and Intel chips helped the machine rank 135th fastest in the world in 2019. In the past two years the People's Armed Police and Public Security Bureau have built regional data centers next door, likely to cut latency as it crunches huge reams of surveillance data. …
The supercomputer center is as bleak a symbol of dystopian tech as you can imagine. It sits at the end of a forlorn road that passes six prisons. The machines, powered by Intel and Nvidia, line the inside of a strange oval-shaped building with an inexplicably green lawn. https://…
@nvidia has deleted this 2017 blog post in which it boasted about how its Tesla GPUs help Huawei, Hikvision, and Dahua superpower their video surveillance analysis. All 3 firms deeply involved in Xinjiang & now sanctioned by US gov. here's the post https://web.archive.org/...
a year ago, @ipvideo reported that @nvidia chips also power PRC police security camera networks that track whether someone is Uyghur by secretly analyzing their facial characteristics https://ipvm.com/... this was not just in Xinjiang, but all across China
We know China uses technology to aggressively monitor its citizens, especially Muslims, but now it merges the systems are powered by US tech Even Nvidia and Intel weren't aware and their chips are used Important reporting from @paulmozur: https://www.nytimes.com/... https://twitt…
As Chinese officials hung thousands of cameras across Xinjiang, an abiding question has been how they process all that footage. We found an answer. They're using one of the world's fastest supercomputers. And it was built with American microchips. https://www.nytimes.com/...
On techmeme. more people should be talking (or at least tweeting) about this besides @paulmozur himself. Nvidia, Intel et al's questionable supply connection with Sugon 中.曙光. https://www.nytimes.com/... https://twitter.com/...
Intel and Nvidia chips power a supercomputing center in China that tracks people in a place where government suppresses minorities, raising questions about the tech industry's responsibility - The New York Times https://www.nytimes.com/...