What threat do Broadcom and Marvell pose to Nvidia?

Broadcom and Marvell offer custom chips (ASICs) to major tech companies, reducing their reliance on Nvidia's GPUs for tasks like inference.

Why are big tech companies moving to custom chips?

To reduce costs and improve performance for specific tasks like inference, rather than relying on Nvidia's general-purpose chips.

Will Nvidia lose its dominance in the AI chip market?

It's still a long way off, but the shift toward custom chips could open opportunities for competitors and reduce Nvidia's share in the inference segment.

Which companies benefit from this shift?

Broadcom and Marvell are the main beneficiaries, along with the big tech companies that design their own chips.

Does this affect Nvidia's stock?

It could pressure Nvidia's stock in the long term if the shift to custom chips accelerates, but Nvidia still dominates AI training.

Broadcom and Marvell: A Silent Threat to Nvidia's AI Chip Dominance

New analysis indicates that Broadcom and Marvell may be a bigger threat than investors realize to Nvidia's dominance in AI chips, as hyperscalers increasingly turn to custom silicon for their growing needs.

According to a report from 24/7 Wall St., Nvidia's (NVDA) dominance in AI training faces a silent threat from Broadcom (AVGO) and Marvell (MRVL). This comes as hyperscalers like Microsoft (MSFT), Amazon (AMZN), Meta (META), and Alphabet (GOOGL) increasingly design their own chips (ASICs) to reduce reliance on Nvidia and cut costs.

Details

The report highlights a strategic shift in how tech giants will spend their next trillion dollars. Instead of buying GPUs from Nvidia, many of these large customers are opting for custom chips tailored to specific tasks like inference or specialized applications. This is where Broadcom and Marvell come in, as they have long-standing expertise in custom chip design (ASICs) and networking silicon.

Broadcom, for instance, has collaborated with Google on Tensor Processing Units (TPUs) and with Amazon on Graviton and Trainium chips. Marvell, meanwhile, supplies high-speed networking chips and custom accelerators to major tech companies.

Context

Nvidia still dominates AI training, but demand for inference chips is growing rapidly. Inference—running a trained model—is often more efficient on custom chips than on general-purpose GPUs. This shift could open significant opportunities for Broadcom and Marvell.

What This Means for Investors

For investors, this suggests the AI chip market may not remain Nvidia's alone. Broadcom and Marvell offer attractive alternatives for large customers, potentially diversifying revenue streams and reducing risk. However, it will take time before these companies seriously threaten Nvidia's dominance.

Broadcom and Marvell: A Silent Threat to Nvidia's AI Chip Dominance

Details

Context

What This Means for Investors

Frequently Asked Questions