AMD and Broadcom Expand AI Infrastructure to Challenge Nvidia
AMD and Broadcom are deploying new chips and infrastructure partnerships with OpenAI and Microsoft to break Nvidia's dominance in the artificial intelligence market.
Advanced Micro Devices (AMD) and Broadcom Inc. are expanding their AI infrastructure capabilities to compete with Nvidia's market dominance. AMD is shifting toward an end-to-end platform comprising silicon, high-speed interconnects, and software tools, with the goal of increasing data center revenue from $16 billion to nearly $100 billion by 2030. To achieve this, AMD plans to launch the Helios rack server with Instinct MI450 GPUs in Q3 2026 and the MI500 series in 2027.
Strategic partnerships are central to these efforts. OpenAI has taken a 10% stake in AMD and will deploy 6 gigawatts of its GPUs, while Microsoft is developing toolkits to convert Nvidia's CUDA models to AMD's ROCm code to reduce vendor lock-in. Broadcom is focusing on custom application-specific integrated circuits and networking, securing a 10-gigawatt deal with OpenAI and projecting a $60 billion to $90 billion market opportunity from its top three custom silicon customers by 2027.
AMD has also solidified its role in high-performance computing, powering 40% of the world's ten fastest supercomputers. This includes the top-ranked El Capitan and the exascale Frontier system. In 2026, AMD will deploy the Lux and Discovery supercomputers at Oak Ridge National Laboratory to support U.S. research in biology and clean energy. The company further expanded its reach via fifth-generation EPYC processors used by Amazon Web Services, Microsoft Azure, and Google Cloud.