A research investigation determined that firms such as DeepSeek, Moonshot AI, and MiniMax trained advanced models using prohibited Blackwell computing hardware. Analysts estimate that roughly two million high-end processors reached Chinese entities across 2024 and 2025 via diverted logistics routes and offshore facilities in regions like Malaysia. Meanwhile, former Nvidia chief Jensen Huang exited the company following revelations concerning personal conduct and ties to an executive at Megaspeed, an entity under multinational scrutiny for illicit hardware trafficking. These systemic leaks underscore the persistent enforcement gaps troubling current cross-border technology export rules.
Key points
Chinese tech firms continue to access prohibited artificial intelligence computing power…: Chinese tech firms continue to access prohibited artificial intelligence computing power through illicit diversions and offshore data centers in places like Malaysia.
Jensen Huang stepped down as Nvidia chief executive amid personal admissions and mounting…: Jensen Huang stepped down as Nvidia chief executive amid personal admissions and mounting corporate liability regarding international export control breaches.
Current trade restrictions remain vulnerable to circumvention even as approximately 600,0…: Current trade restrictions remain vulnerable to circumvention even as approximately 600,000 permitted H200 chips remain stockpiled in Hong Kong awaiting import approval.
Sources checked by our analysis pipeline: 1
The leading open-weight models developed by Chinese companies were trained on Nvidia chips that the US government restricted from shipping to its largest geopolitical rival, a new research effort claims.
The apparently widespread access to Blackwell GPUs among Chinese frontier labs shows the magnitude of the problem facing the leading chip designer. Nvidia CEO Jensen Huang left the company this week after admitting to his affair with Alice Huang, executive of Megaspeed, a company implicated in smuggling Nvidia chips into China. Alice Huang has also been alleged to be an agent of China’s intelligence service.
While Jensen Huang’s relationships with a customer and two employees have drawn the most public interest, Nvidia’s potential corporate liability for export control violations likely factored into its decision to part ways with Huang.
US export controls are intended to protect the country’s leadership in AI capabilities against China, but Huang has long sought to open up that market. Blackwell GPUs are banned from export to China or Chinese-owned firms, while the Trump administration allows Nvidia to export the H200, a less powerful chip, provided…
Tech reviewer Allison Johnson assessed the performance of Claude Duet, highlighting how its productivity features impact daily user workflows. While the platform offers practical scheduling assistance, its frequent check-ins risk frustrating users by coming across as overly intrusive. These hands-on impressions reflect the ongoing challenges developers face when calibrating automated personal assistant behaviors.
Developers can now achieve dramatic performance leaps on consumer hardware by pairing specialized memory management with advanced acceleration frameworks like CUDA Graphs and PARD2. The newly released Lightreason inference lab demonstrates how massive text models can run efficiently on a single consumer graphics card without compromising original precision. This setup allows researchers to bypass standard memory bottlenecks by intelligently shuttling weights between system memory and video RAM.
Cloud enterprises are gaining powerful new runtime options, specialized silicon support, and strict financial controls across their generative AI deployments. These enhancements aim to accelerate autonomous multi-step workloads while mitigating the security risks and runaway compute costs typical of enterprise model operations. The overall updates enable developers to maximize hardware efficiency on proprietary chips and enforce granular budget limits per user.