In the latest forecast by Gartner, global semiconductor revenue is expected to experience a notable rebound, projecting a growth of 16.8% in 2024, reaching a total of $624 billion.
This follows a forecasted decline of 10.9% in 2023, with the market value reaching $534 billion.
The decline in 2023 is attributed to reduced demand from smartphone and PC customers, along with weakness in data center/hyperscaler spending. Alan Priestley, VP Analyst at Gartner, emphasized that despite strong demand for chips supporting artificial intelligence (AI) workloads like graphics processing units (GPUs), it won’t be sufficient to prevent a double-digit decline in revenue for the semiconductor industry in 2023.
However, 2024 is anticipated to be a bounce-back year, with revenue for all chip types set to grow. This growth will be primarily driven by double-digit expansion in the memory market.
The worldwide memory market is expected to face a 38.8% decline in 2023 but is projected to rebound strongly in 2024 with a growth of 66.3%. In particular, NAND flash revenue is forecasted to decline by 38.8% in 2023, reaching $35.4 billion due to anaemic demand and declining pricing caused by oversupply. Gartner analysts foresee a robust recovery in 2024, with NAND flash revenue growing to $53 billion, a 49.6% year-over-year increase.
DRAM vendors, facing oversupply and reduced demand, are witnessing a market price downturn. This has prompted them to reduce inventory, and Gartner analysts anticipate a pricing rebound in the fourth quarter of 2023. The full impact of these pricing increases is expected in 2024, with DRAM revenue forecasted to increase by 88%, totaling $87.4 billion.
The integration of AI techniques, specifically generative AI (GenAI) and large language models, is driving demand for high-performance GPU-based servers and accelerator cards in data centers. This surge in demand is creating a need for workload accelerators to support both training and inference of AI workloads in data center servers. Gartner estimates that by 2027, over 20% of new servers will integrate AI techniques into data center applications.
