Nvidia vs. Broadcom: Nvidia named better buy on valuation

The article frames the two companies as moving beyond their traditional split: Nvidia as a general-purpose GPU designer and Broadcom as a custom silicon maker. Both have since expanded into broader roles across the data center. Nvidia's Grace CPUs have generated more than $5 billion in sales over the last year, and its recent acquisition of Hugging Face for $12.9 billion points to a growing software layer through CUDA, Nemotron, Cosmos, and Isaac. Broadcom, meanwhile, draws 73% of its $16.7 billion in AI semiconductor sales from XPUs, while its infrastructure software brought in $8.8 billion, up 29% year over year, helped by the VMware acquisition.

Customer commitments diverge in shape. Nvidia expanded its relationship with Amazon Web Services to include another 2 million GPUs through fiscal 2029 and is helping fund a 12-gigawatt data center campus OpenAI is building in Ohio alongside SoftBank Energy. Its non-hyperscaler customer book grew 138% year over year to $40 billion. Broadcom's pipeline is more concentrated but almost as large in scope: Anthropic is on a path from 1 gigawatt of Ironwood this year to 5 gigawatts of TPU v8i in 2027, while OpenAI is set to deploy 1.3 gigawatts of Jalapeno inference chips in 2027 and more than 5 gigawatts in 2028. Google Cloud remains a multiyear TPU design partner, and Meta has three generations of custom MTIA chips booked through 2027.

The investment debate appears to hinge on which quality matters more. Broadcom's AI growth rate is faster on paper, with AI revenue expected to double for two consecutive years, but Nvidia's customer breadth and cash generation give it room to secure power and infrastructure years ahead, as CFO Colette Kress described a supply-constrained outlook. The article's conclusion leans on valuation as the tiebreaker: Nvidia's roughly 23 times forward earnings against Broadcom's about 31. Whether that gap holds will depend on how much of Broadcom's custom-chip momentum converts into revenue on schedule, and whether Nvidia's supply constraints ease enough to meet its own guidance.