Nvidia projects US$1 trillion inference market, taps Samsung for LPU chip
At its GTC 2026 conference, Nvidia CEO Jensen Huang projected a US$1 trillion inference market and announced Samsung will manufacture its new LPU accelerator. High-bandwidth memory suppliers SK Hynix, Samsung and Micron are competing to supply HBM for Nvidia’s AI platforms.
1. Nvidia Projects Trillion-Dollar Inference Market
At GTC 2026, CEO Jensen Huang forecast that the global inference chip market will reach US$1 trillion over the coming years, driven by expanding AI applications in data centers, automotive and edge computing, signaling significant revenue potential for Nvidia’s accelerator business.
2. Samsung to Manufacture Nvidia’s LPU Accelerator
Nvidia announced that its new Learning Processing Unit (LPU) will be produced by Samsung Foundry, leveraging Samsung’s advanced process technologies to optimize performance and yield, with volume production expected to begin in late 2026.
3. HBM Suppliers Compete for Nvidia’s AI Memory Contracts
SK Hynix, Samsung and Micron are vying to supply high-bandwidth memory (HBM) for Nvidia’s next-generation GPUs and LPUs, competing on bandwidth, power efficiency and pricing terms, which could influence Nvidia’s unit costs and supply chain resilience.