All articles tagged with #ai gpus
AMD's agreement to give the US government 15% of its AI GPU sales in China is considered largely immaterial to its overall financial health by Citi, as the impact is mainly on low-margin products. The company's main growth drivers remain its mainstream AI GPUs, with strong sales forecasts and key customers like Amazon and Meta. Wall Street maintains a moderate buy rating with a slight upside potential.
AMD has launched the Instinct MI350 series of HPC/AI GPUs featuring a new 3nm process, 185 billion transistors, up to 288 GB HBM3e memory, and support for FP4 and FP6 data types, offering significant performance improvements for AI workloads and competitive metrics against NVIDIA's offerings.
Dell has revealed details about Nvidia's upcoming AI GPUs, codenamed Blackwell, which are expected to consume up to 1000 Watts, a 40% increase in power over the prior generation. This increase in power consumption will require innovative cooling solutions from server makers like Dell. The new GPUs may feature a dual-die design to manage the heat generated, and they are anticipated to be built on a 3nm-class process technology. Dell's CFO highlighted the company's engineering expertise in liquid cooling and power management to prepare for the computational capacity of these next-generation GPUs, which are expected to be available next year.
NVIDIA sold nearly half a million AI GPUs, including the Hopper H100 and Ampere A100, in Q3 2023, with major tech firms like Microsoft and Meta being the top buyers. The strong sales performance positions NVIDIA as a dominant player in the market, and their plans to sell millions of GPUs in 2024 seem achievable. Chinese markets accounted for around 30% of NVIDIA's orders, but new US export restrictions may impact order fulfillment and revenue. Despite this, Omdia predicts that NVIDIA's quarterly sales will continue to increase, driven by high demand for their AI GPUs. NVIDIA aims to capitalize on the growing server industry and improve their AI offerings through next-gen semiconductor and memory standards.