Broadcom vs Nvidia: Custom AI Chips Shift Investment Case

The article compares two chip giants after their latest earnings. Broadcom focuses on custom ASICs, which are purpose-built for specific AI tasks, saving cost by omitting unnecessary features. Nvidia's GPUs are flexible but may waste capabilities when used for narrow workloads. Broadcom has recently been designing these chips for hyperscalers and AI labs, and orders are starting to ramp.

The investment thesis hinges on market share dynamics. The article suggests that hyperscalers may favor data centers with a higher share of custom chips, putting Nvidia at greater risk of losing data center chip share to custom designs than Broadcom losing share to Nvidia. This is because custom chips can be more efficient for specific AI workloads, and Broadcom's partners are already ramping orders.

For investors, the key uncertainty is how quickly hyperscalers adopt custom chips at scale. If Broadcom's partnerships translate into significant order volumes, it could validate the 'hunter' advantage. Conversely, if Nvidia's GPUs remain indispensable, the risk to its market share may be limited. The article currently leans toward Broadcom as the better buy, but the future is still unfolding.