Nvidia, the world’s leading maker of graphics and compute processors, is expected to sell more than 550,000 of its latest H100 compute GPUs in 2023, according to sources cited by Financial Times.

The H100 GPUs are designed for artificial intelligence (AI) and high-performance computing (HPC) applications, and are based on Nvidia’s Hopper architecture.

The H100 GPUs are in high demand by tech firms in the US and other regions, as they offer superior performance and efficiency for various AI and HPC workloads. The H100 GPUs come in different models and configurations, with prices ranging from $30,000 to over $70,000 per card. Nvidia also sells servers that integrate its H100 GPUs with its Grace processors, which are optimized for AI.

Nvidia’s datacenter business, which includes its compute GPUs, generated $4.284 billion in revenue in the first quarter of its fiscal year 2024. The company is likely to see further growth in this segment, as the AI boom drives the demand for its products. However, Nvidia also faces supply constraints from its manufacturing partner TSMC, which is struggling to meet the demand for its advanced chip packaging technology.

Nvidia did not comment on the report, which is understandable given the FTC rules. The company is set to announce its second-quarter results later this month.

Nvidia and MediaTek Partner to Develop Advanced Vehicle Infotainment Systems