Broadcom Stock Price Just Slipped Below $400 — And the Reason Isn’t What You Think

These days, a Broadcom Monday has a specific choreography. Pre-market traders keep an eye out for overnight stories from Menlo Park or Mountain View. Depending on which hyperscaler revealed information about which chip partnership, the stock either lurches or drifts by midmorning. The news on Monday focused more on Broadcom’s orbit than on the company itself. There are rumors that Alphabet is discreetly considering expanding its partnership with Marvell Technology, possibly adding a second supplier of custom AI chips. That was sufficient to break a nine-session winning run and push AVGO below $400, closing at $399.63, down 1.70%.

Over the weekend, Broadcom’s actual business remained unchanged. The Q1 fiscal 2026 figures, which were released in March, continue to be the cleanest chip earnings available. $19.31 billion in revenue, a 29% increase from the previous year. AI-related sales reached $8.4 billion for the quarter, more than doubling from the same period last year, and net income increased by 34%. According to the company’s own disclosure, its backlog for custom AI chips currently stands at about $110 billion. A semiconductor analyst in 2020 would have thought you were describing Intel at its best if you had shown them those figures. The scale at which Broadcom operates is more than twice that.

In essence, the numbers tell a tale about six clients. Instead of general-purpose GPUs like Nvidia’s, Broadcom creates custom AI accelerators, which are chips designed for particular workloads at particular data centers. The biggest known partnerships are with Meta Platforms (co-developing 2-nanometer MTIA chips and Ethernet networking through 2029), Alphabet’s Google (where Broadcom co-designs the TPU line through at least 2031), and, more recently, Anthropic and OpenAI. Gigawatts of compute capacity, not dollars, are used to measure each of those contracts. AI infrastructure is currently being discussed at that scale.

That’s precisely what makes the Marvell rumor unsettling. As this develops, it seems that the market has always viewed Broadcom’s AI business as a collection of embedded partnerships that are intricate, exclusive, and technically challenging. That argument doesn’t change if Google actually hires Marvell to design a portion of its next-generation custom chips, but it becomes less compelling. Broadcom will still manufacture and ship the bulk of Google’s TPU roadmap through 2031. However, there is a slight softening of the implicit monopoly narrative that justifies a price-to-earnings ratio of almost 78. When that softening appears in the headlines, investors are justified in taking some profit off the top of a nine-session rally.

Over the past ten years, Broadcom’s CEO, Hock Tan, has developed this business in a unique manner. VMware, CA Technologies, and Symantec’s enterprise division are examples of acquisitions that were pieced together with strict operational discipline. Over the last five years, AVGO’s market capitalization has increased by about 1,082%, from roughly $141 billion to $1.89 trillion. That run has been one of the most durable wealth creators on the Nasdaq, and it has created an awkward problem: the stock is now expensive by almost every traditional measure. Normalized P/E is set by Morningstar at almost 55. Price-to-sales ratio greater than 28. On most days, the Relative Strength Index flirts with overbought territory, hovering around 70. Although one-year projections still range from $475 to $630, analysts have established consensus price targets that suggest little immediate upside.

The trajectory of AI revenue itself is what sustains the bull case. By the end of 2027, Hock Tan is expected to generate over $100 billion in revenue from custom AI chips. The initial capacity of one gigawatt will be exceeded by Meta’s MTIA deployment. Broadcom’s co-designed TPU capacity will be crucial to Anthropic’s 2027 launch. AI now accounts for roughly half of Broadcom’s business, up from less than 30% a year ago, according to the second-quarter guide, which shows $22 billion in total revenue with about $10.7 billion attributable to AI. The current roadmap does not clearly indicate when that growth will plateau.

When you speak with managers of growth-equity funds in Menlo Park or San Francisco, the pushback is always the same: customer concentration. A few trillion-dollar counterparties, each with the engineering know-how to try something different if they so choose, are the foundation of Broadcom’s AI engine. Because of this, even though the numbers remained unchanged, Monday’s 1.7% decline felt significant. The next earnings report, which is scheduled for June, is therefore more important than the chart indicates. For the time being, Broadcom is in that unique position where the story keeps catching up despite the fundamentals outpacing it.