Broadcom is growing AI chip revenue at extraordinary rates—143% year-over-year in Q2, with guidance for more than 200% growth this quarter—yet its stock sits 24% below its 52-week high. The company has just launched Jalapeño, a custom inference chip built with OpenAI to run AI models more cheaply than general-purpose processors. Investors are concerned about customer concentration (Broadcom's business relies heavily on six large AI firms) and pricing pressure on custom silicon, despite strong underlying demand signals.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Broadcom reported Q2 AI semiconductor revenue of $10.8 billion（約1.7兆円）, up 143% year-over-year, and guided AI chip revenue to grow more than 200% year-over-year this quarter to $16.0 billion（約2.6兆円）. The company also unveiled Jalapeño, a custom inference processor co-developed with OpenAI, designed to run AI models for users rather than train them.
Why it matters
Broadcom supplies six core custom-chip customers including OpenAI, Anthropic, Meta Platforms, and Google parent Alphabet. Custom application-specific chips like Jalapeño are cheaper for customers to operate than general-purpose processors, which makes them increasingly attractive—but also limits what Broadcom can charge for them. The business is growing faster than most large companies, yet shares trade about 24% below their 52-week high, largely because investors worry about concentration risk (dependence on a small customer list) and the pricing pressure inherent in custom silicon.
What to watch
Broadcom plans to deploy racks of OpenAI-designed Jalapeño chips starting late this year, building toward systems that would ultimately draw 10 gigawatts of power. On forward earnings, the stock trades at about 19 times earnings, boosted by revenue growth guided to exceed 80%. Risks include material slowdown in AI revenue growth from the guided pace or a pullback by one of the six anchor customers.
Broadcom's growth trajectory in AI semiconductors is striking by any standard. The company is scaling from $10.8 billion（約1.7兆円） in Q2 AI revenue to guidance of $16.0 billion（約2.6兆円） just one quarter out—a more than 200% year-over-year jump—while companywide revenue is guided to grow 84%. For a company of Broadcom's size, these growth rates have almost no precedent outside the AI build-out itself. Yet the stock's 24% discount from its 52-week high reflects a specific structural concern: the business increasingly depends on a handful of giant customers. If any of those six anchor customers materially cuts spending, or if the overall AI trade reprices, Broadcom's multiple and growth could contract sharply.
The Jalapeño launch illustrates the logic of that customer concentration. Broadcom and OpenAI jointly engineered an inference processor (the step where an AI produces an answer) tailored to OpenAI's workload. The chip is cheaper for OpenAI to run than buying off-the-shelf graphics processors—which is why custom silicon appeals to hyperscalers—but that same cost advantage limits how much Broadcom can charge. As more AI workloads move to application-specific chips, the market expands, yet sellers' pricing power shrinks. The article notes this is arguably the fastest-growing niche in semiconductors, precisely because customers see a clear economic win, but it is not a market where Broadcom can command premium margins indefinitely.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
