NVIDIA, AMD, Broadcom vie for AI inference leadership in 2026

DataM Intelligence has released an analysis comparing the strategies of NVIDIA, AMD, and Broadcom in the AI inference market through 2026. The report indicates that the AI inference infrastructure is entering a more demanding phase, with competition increasingly focusing on overall system efficiency, including power consumption, data transfer speeds, and cost per token.

NVIDIA currently holds the strongest position in the general-purpose commercial AI inference market. However, Broadcom is emerging as a serious competitor, particularly in custom silicon and networking solutions for hyperscale clients. AMD is developing an open alternative, supported by system-level solutions and strategic acquisitions.

The global AI chip market is projected for substantial growth. DataM Intelligence forecasts the market to reach $361.14 billion by 2035, with a compound annual growth rate of 16.64% between 2026 and 2035.

The analysis emphasizes that AI inference economics are a system problem. Raw accelerator speed is becoming less critical than factors like memory bottlenecks, network performance, and efficient orchestration, which directly impact the cost-effectiveness of expensive hardware. Consequently, companies are increasingly focusing on comprehensive solutions that integrate chips, networks, and software.