Nvidia cheaper than Broadcom on next-year earnings, despite ASIC upside

Nvidia and Broadcom occupy different niches in the AI chip market, but their competitive dynamics hinge on whether custom chips can displace Nvidia's entrenched dominance. Nvidia's GPU ecosystem—including software and networking solutions—has become the industry standard for AI training and inference across every major hyperscaler. However, Broadcom's ASIC strategy exploits a real inefficiency: most AI applications waste the GPU's wide capabilities, so custom chips optimized for narrow workloads offer both performance and cost advantages. Alphabet's success selling TPUs directly to clients proves the concept works in practice.

Broadcom's expected 2027 revenue jump to over $100 billion from roughly $40 billion annualized (Q2 basis) reflects confidence that multiple AI companies will move their ASIC orders into production. Yet this growth forecast embeds significant execution assumptions. The valuation analysis reveals the tension: Broadcom's forward P/E looks expensive, but when adjusted for 2027 earnings estimates, it narrows the gap with Nvidia—though Nvidia still trades cheaper. Crucially, analysts expect Nvidia to grow at 43% next year, which the article suggests may be a conservative projection relative to historical performance, leaving Nvidia more room for upside surprise. Broadcom, by contrast, is priced for full realization of client commitments, creating downside risk if any large order slips or is cancelled.