Yazılar

Two Chinese Nationals in California Charged with Illegal Export of Nvidia AI Chips to China

Two Chinese nationals residing in California, Chuan Geng and Shiwei Yang, have been arrested and charged by the U.S. Justice Department for illegally exporting tens of millions of dollars’ worth of advanced Nvidia AI chips, including H100 processors, to China without the required U.S. Commerce Department licenses. The alleged activities took place between October 2022 and July 2025.

Geng, 28, from Pasadena, and Yang, 28, from El Monte, ran ALX Solutions, a company founded in 2022 shortly after the U.S. imposed stringent export controls to curb China’s military modernization by restricting advanced technology exports. ALX allegedly shipped over 20 consignments through Singapore and Malaysia, common transshipment hubs used to conceal shipments destined for China. Payments for these shipments came directly from companies in China and Hong Kong, not the freight forwarders.

From at least August 2023 to July 2024, ALX reportedly purchased more than 200 Nvidia H100 chips—highly advanced AI processors used in training large language models—from San Jose-based Super Micro Computer. Invoices falsely declared end customers in Singapore and Japan, but U.S. export officials could not verify the shipments or companies at the listed locations.

Nvidia emphasized its strict compliance with U.S. export controls, noting that diverted products receive no support or updates. Super Micro affirmed its commitment to regulatory compliance and cooperation with authorities.

Geng was released on $250,000 bond following a federal court appearance in Los Angeles, while Yang, who overstayed her visa, faces a detention hearing on August 12.

U.S. Senators Call for Probe into Data Security Risks of Chinese AI Model DeepSeek

A group of seven Republican U.S. senators led by Ted Budd urged the Commerce Department on Tuesday to investigate potential data security risks associated with Chinese open-source AI models such as DeepSeek.

The senators—including Jon Husted, Todd Young, John Cornyn, John Curtis, Bill Cassidy, and Marsha Blackburn—requested an assessment of whether applications using DeepSeek collect data that is transmitted back to Chinese servers, and if these AI models are sharing American personal or corporate information with China’s military or military-linked companies.

Their letter also sought information on any improper access by Chinese open-source models to export-controlled semiconductors or breaches of usage terms of U.S. AI models aimed at enhancing Chinese AI capabilities.

Bipartisan legislation has been proposed to ban DeepSeek’s use on federal government devices and networks, as well as prohibit its use by federal contractors in government projects.

Commerce Secretary Howard Lutnick stated in January that DeepSeek appeared to have misappropriated U.S. AI technology and promised to enforce restrictions. The Commerce Department did not immediately respond to requests for comment.

In June, Reuters reported that DeepSeek was assisting China’s military and intelligence services and was attempting to use Southeast Asian shell companies to obtain advanced semiconductors barred from shipment to China under U.S. export rules.

These developments underscore growing skepticism in Washington over DeepSeek’s rapid rise, with officials suggesting the Chinese firm’s AI prowess heavily depends on U.S. technology.

Based in Hangzhou, DeepSeek shocked the tech world in January by claiming its AI reasoning models matched or outperformed leading U.S. models at a fraction of the cost.

Samsung Profit Plunges 56% Amid AI Chip Woes, U.S. Export Curbs to China

Samsung Electronics reported a steep 56% year-on-year drop in Q2 operating profit, projecting earnings of 4.6 trillion won ($3.36 billion)—significantly below analyst expectations of 6.2 trillion won, according to LSEG SmartEstimate. This marks Samsung’s weakest quarterly performance in six quarters, as its semiconductor division continues to struggle with shifting global dynamics in the AI chip market.

The South Korean tech giant blamed its sharp decline on U.S. restrictions on AI chip exports to China, which have disrupted its sales pipeline. However, analysts pointed to delays in delivering high-bandwidth memory (HBM) chips to Nvidia as a major factor in its underperformance. Unlike rivals SK Hynix and Micron, which have seen strong AI-driven chip demand, Samsung has been slower to supply its latest HBM3E 12-layer chips, with customer evaluations still ongoing and no specific update on Nvidia shipments.

“Everything ultimately comes back to HBM,” said Ryu Young-ho, an analyst at NH Investment & Securities, noting that Samsung’s competitive edge hinges on reclaiming leadership in the HBM segment.

Revenue for the quarter is expected to come in nearly flat at 74 trillion won, down just 0.1% from a year ago. But the semiconductor division likely took the hardest hit, with analysts estimating its operating profit may have dropped over 90% to just 500 billion won, partly due to inventory value adjustments and unsold HBM stockpiles.

Adding to the challenges, potential U.S. tariffs and mounting competition in China—where Samsung still has a heavy market presence—are expected to weigh on both its chip and smartphone margins in the near term.

Samsung’s foundry business also saw falling earnings, attributed to low utilisation rates and inventory write-downs, stemming from the same U.S. AI chip export restrictions. However, the company expects foundry performance to gradually improve in the second half of 2025 as utilisation recovers with demand.

Despite the weak outlook, Samsung announced a 3.9 trillion won ($2.85 billion) share buyback, part of a broader 10 trillion won repurchase plan unveiled in late 2024. Investors remained cautious, with Samsung shares slipping 0.2%, while Korea’s benchmark KOSPI index rose 1.2% during morning trading.

Looking ahead, Samsung hopes to recover with upcoming phone launches and by expanding HBM sales beyond Nvidia to other customers. A full breakdown of business unit performance is expected on July 31, when the company releases its detailed Q2 earnings report.