Broadcom vs. Nvidia: Is Custom Silicon the Next Big Thing in AI?

Two finance podcast hosts recently debated where AI value will accrue next. One host argued for Broadcom, citing its custom accelerator business and impressive growth. In Q3 of fiscal 2026, Broadcom's AI semiconductor revenue rose 221% year-over-year to $16.7 billion. The company guided the fourth quarter to $21.7 billion in revenue, up 236%. Hock Tan sketched a trajectory to roughly $115 billion in fiscal 2027 and $230 billion in fiscal 2028.

The host also mentioned that OpenAI's latest release required 100 Vera Rubin chips to power it, highlighting the importance of custom silicon for stable workloads. Broadcom builds application-specific integrated circuits (ASICs) for hyperscalers like Google and Meta, while Nvidia sells one platform to everyone.

Taiwan Semiconductor is another player in the space, fabricating both Broadcom's XPUs and AMD's Instinct GPUs. Dell Technologies booked a record $60.9 billion in AI orders last quarter with a $95 billion backlog, raising its full-year FY27 guide by $25 billion to $192 billion.

Broadcom trades at a trailing P/E of 46x and a forward P/E of 19x, with an analyst target of $531.85 against a recent price of $361.99. Shares are down 12.99% over the past month, even after a Q3 that logged nine consecutive EPS beats and free cash flow of $13.66 billion.

The risk/reward for Broadcom looks constructive, with eyes open. The company's custom-accelerator franchise is priced as though the $115 billion fiscal 2027 outlook might not arrive, but the setup rewards the patient reader willing to sit through the noise.