Broadcom’s Real Risk Is Its Customer List
Nearly half of Broadcom’s revenue now comes from AI semiconductors, and the accelerators inside that line are built for a short list of core customers.
Broadcom (AVGO) trades near $356, about 26% below its 52-week high, and the easy read on an AI winner down that far is that the stock has finally gone on sale. It has not. The number a holder should sit with is not the size of the decline; it is how much of this company now rides on the purchase plans of a very short list of buyers.
AI Semiconductors Are 49% Of Revenue, And The Fiscal Q3 Guide Pushes That Higher
In fiscal Q2 2026, AI semiconductors accounted for 49% of quarterly revenue of $22.2 billion, by the company’s own breakdown. That is no longer a growth business bolted onto a diversified chip and software franchise; it is the franchise. The fiscal Q3 2026 guide is heavier still: $16 billion of AI semiconductor revenue inside $29.4 billion of consolidated revenue, roughly 54% of the total. That consolidated figure is itself a sharp step up from the $22.2 billion just reported. Broadcom is not diversifying into AI so much as converting into it.
Six Core Customers Carry The Accelerator Business
The custom accelerators inside Broadcom’s AI semiconductor line are built for six core customers. One of the six holds a long-term agreement covering multiple generations of TPUs, Broadcom has a contractual commitment to deploy 1.3 gigawatts for OpenAI in 2027, and Meta’s MTIA X deployment is planned through the end of 2028. The order book behind that list is genuinely deep: AI bookings topped $30 billion in fiscal Q2 2026 against $10.8 billion actually shipped, and management put its forward visibility at 2028.
Depth is not breadth, though. Networking, which Broadcom also sells into deployments that do not use its own accelerators, made up almost 40% of the AI line in fiscal Q2 2026, and management expects that share to settle closer to 30%. The broadest part of the AI business is the part set to shrink as a share. Six buyers making six capital decisions is a different exposure from a market of thousands, and management has already conceded the point on the customer behind that TPU agreement: Broadcom fully expects some diversity of sources there. Owning a stock with a revenue base shaped like that is a different proposition from owning the Trefis High Quality Portfolio, which does not depend on the handful of largest technology names to produce its returns.
Down 15% In Three Months And Still Near A Decade-High Sales Multiple
The concentration would matter less if the price left room for it. Down 15.3% over the past three months, the stock still fetches 22.4 times sales, close to the ten-year high of 26.0 and in the top decile of its own decade. At that multiple a buyer pays for today’s growth and today’s margins together, on a revenue base whose fastest-growing slice takes its direction from six boardrooms. Broadcom is executing rather than stumbling, and it stayed ahead of the S&P 500 over the past year. But a holder buying this decline is not buying a discount, and the dip-buyer’s playbook is the honest place to test that before adding. When fiscal Q3 2026 is reported, the figure that matters is not the $16 billion of AI revenue; it is whether the customer list is any longer than six.
Could One Warning Sign Cost You Years Of Gains?
A warning in the fundamentals is manageable when you are diversified and dangerous when you are not. Concentration tends to arrive by accident rather than by decision. What your largest position would do to your net worth is exactly what the Trefis Wealth team computes, with the same rules-based systematic discipline that runs our High Quality Portfolio. Request a free vulnerability audit of your biggest positions.