Nvidia and Broadcom Pivot to AI Data Center Architecture

Broadcom and Nvidia Inc. are evolving from semiconductor specialists into architects for the entire AI data center, reporting massive revenue growth driven by global infrastructure expansion. Nvidia reported second-quarter data center revenue of $89 billion, a 117% year-over-year increase, and is guiding toward 70% growth for fiscal 2028. To expand its software ecosystem, Nvidia acquired Hugging Face for $12.9 billion and is scaling its Vera Rubin CPUs.

Broadcom is seeing rapid growth in custom AI semiconductors, with AI revenue projected to reach $115 billion in fiscal 2027 and $230 billion by fiscal 2028. The company reported a 221% year-over-year increase in AI semiconductor revenue to $16.7 billion for the third quarter of fiscal 2026. While Nvidia provides a broad computing platform for a wide customer base, Broadcom focuses on purpose-built engines for a few giant hyperscalers, including Alphabet and Meta.

A competitive tension has emerged as hyperscalers develop custom accelerators to lower costs and reduce reliance on Nvidia GPUs. OpenAI co-developed the Jalapeño chip with Broadcom specifically to run large AI models more efficiently. To further its infrastructure goals, OpenAI is building a 12-gigawatt data center campus in Ohio with funding assistance from Nvidia and SoftBank Energy.