Jensen Huang Resigns from Nvidia as Smuggling Inquiries and Chip Contests Mount
Jensen Huang has stepped down as Nvidia CEO following personal controversies and an alleged $7.5 billion GPU smuggling operation, while hardware rivals roll out new AI silicon.
Nvidia founder Jensen Huang has stepped down as chief executive officer. The sudden leadership change follows personal controversies and scrutiny surrounding an alleged 7.5 billion dollar artificial intelligence chip smuggling scheme linked to Shanghai fund manager Kang Wei and his wife, Alice Huang.
Export Limits and Smuggling Pressure
Securities regulators in China have targeted Kang Wei, who reportedly helped construct an illicit network moving Nvidia hardware into the country. Despite strict export controlsExport ControlsGovernment rules that limit or forbid the sale of certain sensitive technologies to specific foreign countries. Open in glossary, Chinese artificial intelligence labs continue seeking high-end compute resources. Labs have placed orders for more than 6 million units of the export-compliant RTX 5500 Pro, which uses the GB202 processor. Meanwhile, black-market prices for B300 racks have reached 2.8 million dollars as buyers turn to cloud rentals in Malaysia and Singapore. In response, Nvidia has imposed stricter rules on cloud partners, requiring a 51 percent quota of non-Chinese clients.
Rival Silicon and Local Compute
Competitors are adjusting to ongoing supply limits and shifting market demands. Google expanded its eighth-generation Tensor Processing Unit line by launching the TPU 8i Max and the TPU 8tx. The TPU 8i Max places two chips on a single package for long-context workloads. The TPU 8tx pairs with AMD EPYC host processors to focus on specialized security tasks.
For local artificial intelligence tasks, tests evaluated Nvidia’s DGX Spark against Apple’s Mac Studio running the M5 Ultra processor. System software stacks like MLX are becoming necessary to handle larger local inferenceInferenceRunning a trained AI model to get an answer. Open in glossary workloads on personal workstations.
Memory Solutions and Automated Design
Memory shortages remain a challenge across the entire semiconductor sector. AMD detailed its MEXT technology, a predictive memory system driven by a transformer model and flash storage. MEXT aims to maintain 95 to 99 percent of standard server performance while cutting system dynamic random-access memory needs in half. AMD currently operates 145 servers running MEXT and aims to scale to 2,000 systems next year.
Separately, startup Cognichip unveiled an automated software suite called Artificial Chip Intelligence. The company reports the platform can compress chip design timelines from months to days by handling repetitive engineering tasks, though the tools have not yet appeared in shipping silicon.
How does this story land?
One tap per reader.
Quick poll
Can predictive software like AMD MEXT permanently ease dynamic memory constraints?
One tap. No sign-up.
Ask this story
Still have a question?
Answers use only what is in this story. They are written by AI and can be wrong.