Yole Group released its "Data Center Semiconductor Trends 2025" report, detailing the impact of artificial intelligence (AI), high-performance computing (HPC), and hyperscaler demand on the data center semiconductor market. The report projects the market to grow from $209 billion in 2024 to $492 billion by 2030, driven by generative AI and evolving compute and memory architectures.
In 2024, GPUs generated $100 billion in revenue, with Nvidia holding 93% of the server GPU market. By 2030, GPU revenue is expected to reach $215 billion, fueled by their critical role in AI training and increasing use in inference. AI ASICs are also gaining traction, with hyperscalers like Google, Amazon, and Microsoft investing in custom silicon for performance optimization and cost control. AI ASIC revenue is forecasted to reach $84.5 billion by 2030.
Memory and interconnect solutions are addressing AI-driven bottlenecks. DDR5 adoption is rising, HBM is in high demand for AI training, and CXL is gaining traction to tackle memory disaggregation and latency issues. Photonics technologies, including optical I/O and co-packaged optics (CPO), are also emerging to support interconnect needs.
The competitive landscape remains U.S.-dominated, with Nvidia, AMD, and Intel leading. However, China is expanding its domestic semiconductor capabilities through strategic investments and policies, influenced by export controls. Startups such as Groq, Cerebras, and Tenstorrent are introducing innovative chip designs, challenging established players on cost, performance, and energy efficiency.
The report highlights that AI workloads are reshaping chip design, packaging, and manufacturing, with a focus on resolving compute, memory, and interconnect challenges in data center infrastructure.
For more information, contact Yole Group at https://www.yolegroup.com.
