Nvidia chips

U.S. prosecutors indicted four individuals for conspiring to illegally export millions in Nvidia chips to China and Hong Kong, circumventing U.S. export controls. The scheme allegedly involved routing the chips through Malaysia and Thailand to obscure their final destination, requiring falsified documentation and failure to secure necessary licenses. The defendants face charges related to violating the Export Control Reform Act, smuggling, and money laundering. The case highlights U.S. concerns over China’s AI and supercomputing ambitions and their potential military applications.

The U.S. has approved the sale of up to 35,000 advanced Nvidia chips, worth ~$1 billion, to Saudi Arabia’s HUMAIN and UAE’s G42. This reverses the previous policy restricting AI tech exports to Gulf state-backed entities due to concerns about potential diversion to China. The Trump administration cites “promoting American AI dominance” as the reason, contingent on strict security and reporting requirements. The move coincides with Saudi Arabia’s pledge to invest $1 trillion in the U.S. economy, and HUMAIN’s AI partnerships with major tech firms, including xAI.

Oracle’s stock dipped 3% following a report highlighting lower-than-expected gross margins (14%) in its Nvidia-powered cloud division compared to its overall margin (70%). Despite this, Oracle’s cloud business is experiencing rapid growth, projecting $144 billion in revenue by 2030, fueled by projects like “Stargate” with OpenAI. The report raises concerns about the financial viability of Oracle’s AI cloud strategy given Nvidia GPU costs and competitive pricing, needing careful cost management for long-term profitability.

The US-China AI chip battle escalates as Nvidia’s restricted H200 and B200 GPUs enter China via shadow networks during a 90-day tariff reprieve, fueling a volatile black market. Cloud giants face acute scarcity, while emerging hybrid supply chains disguise GPUs as industrial goods. Structural contradictions emerge: despite speculative bubbles and unviable projects, specialized AI adoption grows, exposing systemic bottlenecks in technical innovation, data readiness, and vertical integration. Government subsidies clash with industry demands for foundational ecosystem reforms as companies pivot to VC-driven compute models amid shifting demand from pre-training to inference workloads.