Nvidia at 14.4 times earnings edges Broadcom as AI pick

Both Broadcom and Nvidia sit on the same global AI infrastructure build-out, but they meet it from different angles. Broadcom helps large customers build custom chips for specific workloads, while Nvidia sells a broader computing platform serving many customers and use cases. That difference shapes how each company's growth looks. Broadcom's AI semiconductor revenue jumped 221% year over year to $16.7 billion, and management now expects AI chip revenue of $115 billion in fiscal 2027 and $230 billion in fiscal 2028. Nvidia is growing from a much larger base: revenue surged 106% year over year to $96.2 billion, with Data Center revenue up 117% to $89 billion.

The competitive thread running through the comparison is custom silicon. Some large AI companies are developing their own chips to cut computing costs and optimize inference, the step where an AI model runs in production. OpenAI's Jalapeño chip, co-developed with Broadcom, is one example; according to Reuters, OpenAI developed it partly to reduce its reliance on Nvidia GPUs and lower operating costs. The threat is most direct among hyperscalers (large cloud providers), which have the scale and resources to design their own chips.

Nvidia's answer so far is breadth. Revenue from its AI cloud, industrial, and enterprise customers jumped 138% year over year in the second quarter, faster than the 102% growth in its hyperscale business, suggesting growth is spreading beyond the buyers most able to design their own silicon. On valuation, Nvidia traded at roughly 14.4 times forward one-year earnings versus about 18.8 times for Broadcom as of Sept. 10. The outcome likely hinges on whether custom accelerators meaningfully dent Nvidia's inference share — and on whether Nvidia's wider customer base keeps growing faster than its largest accounts.