Overview
- Both Broadcom and Marvell have moved from data‑center networking into custom AI ASICs that hyperscalers use to lower cost and boost performance compared with general‑purpose GPUs.
- Broadcom currently runs a much larger custom‑chip business with multiple marquee clients and multibillion‑dollar AI revenue and bookings that give it greater scale and near‑term visibility.
- Marvell has pursued different commercial terms that tie equity upside to customer purchases, a structure that recent filings show can link future payments to tranche‑based buys and that has changed how markets price its growth prospects.
- Investors are valuing the two companies very differently, with Marvell trading at a much higher forward multiple after a steep share rise and Broadcom priced as a larger, lower‑multiple infrastructure supplier.
- The next key drivers will be hyperscaler procurement pacing, each supplier’s design and manufacturing execution, and how binding customer commitments are counted in revenue forecasts and investor models.