Broadcom's Latest Earnings Confirm My Deepest Concerns About This AI Stock

In the AI boom, Broadcom (AVGO) has been viewed by the market as an important beneficiary of computing power, but its latest quarterly earnings have exposed structural concerns behind the glossy growth. I maintain a sell rating on Broadcom, with a target price of approximately $393. Previously I was neutral on Broadcom, while I preferred other AI beneficiaries such as Google, Nvidia, and Micron.

Key Takeaways

• AI semiconductor revenue is growing rapidly, but revenue quality is deteriorating, and gross margin faces sustained downward pressure.
• Customer concentration is high, and key customers such as Google are intensifying pricing negotiations and introducing second sources, weakening Broadcom's bargaining power.
• To lock in orders from large customers, Broadcom has accepted lower gross margins and more aggressive commercial terms on custom chips.
• VMware remains Broadcom's most attractive business segment, but as the share of low-margin AI hardware rises, its overall contribution is being diluted.

AI Growth Masks Declining Gross Margin

On the surface, Broadcom's AI semiconductor revenue growth is very strong, seemingly fully enjoying the benefits of this computing expansion. But upon deeper analysis, this growth relies heavily on custom chips for a few super-large customers, and the gross margin of these custom businesses is significantly lower than the company's traditional products. At the same time, Broadcom has made considerable concessions in commercial terms to compete for and lock in orders, further compressing profit margins. Therefore, the higher the share of AI revenue, the more pressure the company's overall gross margin faces.

Large Customers' Bargaining Power Continues to Rise

Google, one of Broadcom's major customers, is taking a more proactive cost-control approach. On the one hand, Google is pressing for lower unit costs in price negotiations; on the other hand, it is actively seeking and supporting other chip suppliers to reduce its dependence on Broadcom. Even if long-term agreements are in place, this shift in the supply-demand relationship still weakens Broadcom's economic leverage. Once a large customer successfully diversifies its supply, Broadcom's future revenue visibility and profit margins will be challenged.

VMware's Value Is Being Diluted by Low-Margin Hardware

In Broadcom's business portfolio, VMware has always been regarded as the most attractive asset, with stable software revenue and higher profit margins. However, as AI hardware revenue grows rapidly, VMware's share of total revenue is declining. This means Broadcom's revenue structure is tilting toward low-margin, high-capex hardware, and the profit cushion provided by the software business is gradually being diluted. For shareholders, this is not a healthy trend.

Compared with Other AI Beneficiaries

Compared with Broadcom, I prefer companies with stronger pricing power, a more diversified customer base, or higher added value in the AI value chain, such as Google, Nvidia, and Micron. These companies either have monopolistic advantages in their software ecosystems or control scarce capacity in core hardware, enabling them to better convert AI demand into sustainable profit growth. Broadcom's current challenges — large customer price pressure and expansion of low-margin custom business — are exactly the risk signals to be wary of in this AI cycle.

Summary

Broadcom's latest earnings have confirmed my previous concerns: AI-driven revenue growth has not brought profit improvement of the same quality, but instead has intensified the risks of customer concentration and gross margin compression. Against the backdrop of large customers such as Google continuously increasing their bargaining power and VMware's contribution being diluted, the current stock price of about $393 does not fully reflect these structural challenges. Therefore, I maintain a sell rating on Broadcom and recommend that investors prudently assess the true earnings quality of its AI business.

Sources and Editorial Notes

This article is based on the public original article link to distill the theme, and is rewritten and structurally edited without adding unverified facts; the views of the original author do not represent this site, and dynamic data should refer to the original page.