◆ What the Market Isn't Pricing In

Broadcom is not a cheap stock, and any note about it must begin there. After the AI boom re-rated the shares to roughly twenty times sales and forty-six times earnings¹, the market has plainly priced in a great deal of optimism — this is not a misunderstood bargain but a richly-valued business that must keep delivering to justify its price. So the interesting question is not whether the market has noticed Broadcom's AI growth (it emphatically has) but whether it is mispricing the character of that growth and the business wrapped around it.

The first thing the market may misjudge is the quality of Broadcom's AI franchise. It is fashionable to treat AVGO as a lower-grade Nvidia proxy — a picks-and-shovels play riding the same wave — but the two are quite different animals. Broadcom's AI revenue comes largely from custom silicon it co-designs with a handful of hyperscalers and from the networking that stitches their clusters together, and that business is stickier and more visible than selling merchant chips. A custom accelerator is engineered into a customer's system for years; $179.2 billion of committed orders gives a chip company an almost unheard-of view of demand years out². This is design-win revenue, not spot-market revenue, and design-win revenue is worth more.

The second thing easily lost in the AI noise is the software annuity underneath. Broadcom is not a pure semiconductor company; a large share of its revenue is now recurring infrastructure software, VMware above all — high-margin subscriptions from enterprises that cannot easily leave. That annuity smooths the notorious cyclicality of chips and throws off steady cash regardless of the AI weather, which makes Broadcom a fundamentally more stable, higher-quality enterprise than the 'AI chip stock' label implies. The market, transfixed by the AI semiconductor line, tends to undervalue the boring, dependable software cash flows doing the stabilizing.

But even-handedness cuts the other way, and the bear case is equally real. That rich valuation leaves no margin for error: it assumes years of uninterrupted AI growth from a handful of customers who are, uncomfortably, also Broadcom's most capable potential competitors, and who are building their own silicon. The AI spending is a boom, and booms are cyclical; the backlog is a strong signal, not a contract. And the software cash that looks so stabilizing was bought with a price-hike strategy that has provoked a genuine customer revolt, the long-term cost of which is still unknown. Everything that makes Broadcom attractive comes with a corresponding risk the high multiple does not forgive.

So the observation is not that Broadcom is cheap — it isn't — but that the market may be mispricing the shape of the business behind the price: a higher-quality, more-visible, software-stabilized AI franchise than the cyclical-chip framing suggests, offered at a valuation that simultaneously demands that everything go right. The thing the market may not be fully pricing is that Broadcom is both better and riskier than its label — a genuinely durable compounding machine and a richly-valued bet on a boom, at the same time. Which of those two truths dominates from here is the question every owner of these shares is really answering.