The Tech World at a Turning Point: Nvidia Changes and AI Innovations

The technology industry is undergoing significant shifts. Last week, all eyes were on Nvidia as its founder and CEO, Jensen Huang, unexpectedly stepped down. This decision marks a major challenge for a company that has transformed over the last five years from a gaming graphics card manufacturer into a global leader in artificial intelligence. ## Nvidia’s Crisis and Market Shifts Huang’s departure is accompanied by controversial details regarding his personal life and connections to Chinese business circles. Reports suggest the matter involves illegal schemes related to the smuggling of Nvidia graphics processors into China. Simultaneously, Nvidia is implementing strict measures to prevent the supply of AI computing power to China via cloud services in Asian countries. Amidst market shortages, the price of B300 server systems on the black market has reached $2.8 million. ## Local AI and New Technological Standards While the global market monitors the situation surrounding Nvidia, enthusiasts and specialists are increasingly focusing on running AI models locally on personal computers. An interesting rivalry has emerged between Nvidia’s DGX Spark and Apple’s Mac Studio (with the M5 Ultra chip). Testing shows that choosing the right hardware for specific workloads is now critical. Furthermore, AMD has introduced MEXT technology—a predictive memory system that allows servers to maintain 99% performance even when DRAM is halved. This innovation is based on transformer models and the intelligent use of flash memory, which will help the server industry address memory shortages. ## Innovations in Chip Design and Security In parallel, Google has expanded its TPU (Tensor Processing Unit) line with two new chips: the TPU 8i Max and the TPU 8tx. The former focuses on AI agents with long-context capabilities, while the latter utilizes AMD EPYC processors for specialized security tasks. At the same time, Cognichip has announced AI-based tools that will reduce the semiconductor design process from months to days. Finally, to improve AI safety, the Frontier Model Foundation has proposed the “Jailbreak Rating Framework”—a system designed to evaluate the vulnerability of AI models and their potential to generate harmful content. These initiatives aim to foster safer and more efficient technological development.