Broadcom delivered exactly what Wall Street wanted: record revenue, earnings beats, and AI growth that would make any competitor envious. Yet the stock dropped 5% Friday morning. The disconnect reveals something important about where we are in the AI investment cycle.

The Numbers That Should Have Been Bullish

Broadcom's fiscal fourth quarter results were objectively impressive:

  • Revenue: $18.02 billion vs. $17.45 billion expected (up 28% YoY)
  • EPS: $1.95 vs. $1.87 expected
  • AI revenue: Up 74% year-over-year in Q4
  • Full-year AI revenue: $20 billion (up 65% YoY)
  • Q1 2026 guidance: $19.1 billion revenue, AI revenue to double to $8.2 billion

CEO Hock Tan also revealed a $73 billion backlog for custom AI chips, switches, and data center components over the next 18 months. And in a major customer announcement, Tan confirmed that Anthropic—the AI startup behind Claude—has placed an $11 billion order on top of an initial $10 billion commitment.

A fifth custom chip customer with a $1 billion order was also secured.

So why did the stock fall?

The Margin Problem

Buried in the earnings call was a warning that spooked investors: gross margins are expected to compress as AI becomes a larger portion of sales.

The issue is structural. Broadcom's AI business involves custom silicon designed specifically for hyperscale customers. These chips require expensive components—advanced packaging, high-bandwidth memory, specialized materials—that Broadcom passes through to customers without significant markup.

The result: AI revenue grows rapidly, but each AI dollar contributes less profit than traditional semiconductor revenue.

For a company trading at premium multiples, margin compression is a serious concern. Investors have been paying for growth AND profitability. Getting only growth changes the calculus.

The Customer Concentration Question

Broadcom's AI business is built on a handful of massive customers: Google, Meta, reportedly Amazon, and now confirmed Anthropic. Each relationship is worth billions—and each represents concentration risk.

When Tan ended the mystery around the "fourth customer" by naming Anthropic, it was bullish news. An $11 billion order validates demand. But it also highlighted how dependent Broadcom's AI growth is on winning—and keeping—a small number of whale accounts.

What happens if Google decides to design more chips in-house? What if Anthropic's funding situation changes? These aren't immediate threats, but they're risks that justify some valuation discount.

The Comparison to Oracle

Broadcom's selloff came just 24 hours after Oracle crashed 13% on its own AI-related concerns. The back-to-back declines suggest investors are growing more skeptical of AI investments broadly.

But there's an important distinction: Oracle is spending $50 billion hoping to attract AI customers. Broadcom already has them—and $73 billion in backlog to prove it.

The difference between "building AI infrastructure hoping customers come" and "fulfilling orders from customers who have already committed" is significant. Broadcom's situation is fundamentally stronger than Oracle's, even if both stocks are struggling.

What the Dividend Increase Signals

One unambiguously positive element: Broadcom raised its quarterly dividend by 10% to $0.65 per share. CFO Kirsten Spears cited "increased cash flows in fiscal year 2025" as the rationale.

Dividend increases are commitments. Management doesn't raise dividends unless they're confident in sustained cash generation. The hike suggests Broadcom's leadership sees the margin compression as manageable—not existential.

The Bigger Picture: AI Stocks Are Being Repriced

Broadcom's selloff, following Oracle's crash, fits a pattern: the market is demanding more from AI investments.

In 2023 and 2024, simply having AI exposure was enough to drive stock prices higher. Investors bought the promise of future growth.

In late 2025, the bar has risen. Investors want:

  • Revenue growth (Broadcom delivered)
  • Earnings beats (Broadcom delivered)
  • Strong guidance (Broadcom delivered)
  • AND maintained margins (Broadcom struggled here)

This isn't a rejection of the AI thesis. It's a maturation. The easy money has been made. Future returns require picking companies that can grow profitably—not just grow.

What Investors Should Do

Broadcom's 5% decline doesn't invalidate the investment case. The company remains one of the best-positioned AI beneficiaries, with:

  • Confirmed, contracted demand ($73 billion backlog)
  • Diversified customer base (five major custom chip customers)
  • Software revenue from VMware providing stability
  • Commitment to shareholder returns (dividend increase, buybacks)

The stock is now cheaper than it was yesterday. For long-term investors, that may represent opportunity rather than warning.

However, the margin warning is real. Monitor gross margins in coming quarters. If AI revenue mix continues rising while margins compress, the growth story becomes less valuable.

The Bottom Line

Broadcom's Q4 results were strong by any objective measure. AI revenue surging 74%, record bookings, and confirmed mega-orders from top AI companies represent exactly what investors hoped for.

The stock fell anyway because expectations have shifted. In the current market, AI growth alone isn't enough. Profitable AI growth is the new standard.

For Broadcom, the question is whether the company can manage the margin compression inherent in its custom chip model. The $73 billion backlog suggests customers believe the value proposition works. Whether shareholders agree will determine where the stock goes from here.