3Before9

Samsung Electronics will manufacture NVIDIA's new Groq 3 Language Processing Unit, the inference-focused chip that emerged from NVIDIA's USD 20 billion acquisition of Groq last December. Jensen Huang confirmed the partnership on stage at GTC 2026 in San Jose on 16 March, sending Samsung shares up as much as 5.3 per cent. The Groq 3 LPU uses on-chip SRAM rather than traditional high-bandwidth memory, enabling faster data transfer and better power efficiency for inference workloads. Samsung has already ramped wafer production from roughly 9,000 to 15,000 units on its 4-nanometre process, with commercial shipments expected in Q3 2026. The Korean chipmaker also unveiled its seventh-generation HBM4E memory at the same event, doubling down on its position in the AI silicon supply chain.

Four major Asian automakers - China's BYD and Geely, Korea's Hyundai, and Japan's Nissan - have committed to building level 4 autonomous vehicles on NVIDIA's DRIVE Hyperion platform. The announcement, made at GTC 2026, marks the broadest adoption yet of NVIDIA's standardised autonomous driving architecture by Asian manufacturers. Level 4 vehicles can operate without human intervention under defined conditions, a significant step beyond the assisted-driving features currently shipping in most markets. NVIDIA also introduced Halos OS, a safety architecture designed to provide a universal foundation for production-ready autonomy, and an updated Alpamayo 1.5 open model with improved multi-camera support. Separately, mobility platforms including Singapore's Grab confirmed they are leveraging the DRIVE Hyperion stack for their own autonomous vehicle programmes.

South Korean e-commerce giant Coupang has built what it calls an "AI factory" in collaboration with NVIDIA, combining its proprietary Coupang Intelligent Cloud with NVIDIA DGX SuperPOD infrastructure. The company presented results of the partnership at GTC 2026, revealing it has also signed on as a launch partner for NVIDIA's Dynamo, an open-source inference operating system designed to maximise GPU utilisation within data centres. By adopting Dynamo, Coupang expects to increase its data processing throughput by up to 30 times. The additional compute power will be directed at demand forecasting and delivery route optimisation, the twin engines behind Coupang's signature "Rocket Delivery" service that promises next-day or same-day fulfilment across Korea.