Broadcom’s 20% Dip Puts a Spotlight on AI Spending Risk, Even as Earnings Announcement Continued Strength

Broadcom Inc. (AVGO) shares slid about 20% from a recent peak after the company delivered a strong fiscal second-quarter report but issued guidance that some analysts viewed as less upbeat than expected. The sell-off has intensified scrutiny of whether hyperscaler and large AI lab spending on data centers and accelerators will keep accelerating, or whether customers will push back as their costs rise.

Broadcom supplies the custom chips and related infrastructure that power AI training and inference, including data center networking equipment and accelerator designs built for specific customers. The latest coverage notes Broadcom’s role in supporting hyperscalers’ custom Tensor Processing Unit strategies at Alphabet, while also serving AI startups with accelerators intended to expand compute access beyond off-the-shelf GPU-based supply chains.

In its official results, Broadcom reported fiscal Q2 revenue of $22.2 billion for the quarter ended May 3, 2026, up 48% year over year. AI semiconductor revenue came in at $10.8 billion, up 143% year over year, driven by demand for custom AI accelerators and AI networking, according to management. The company also forecast AI semiconductor revenue to increase to $16.0 billion in fiscal Q3, implying growth of more than 200% year over year.

Broadcom’s guidance for the broader quarter also suggested momentum. For fiscal Q3, the company guided consolidated revenue of approximately $29.4 billion, an 84% year-over-year increase, alongside non-GAAP operating income guidance of about 67% of projected revenue and adjusted EBITDA guidance of about 68% of projected revenue. Broadcom also described the quarter as producing record revenue, operating profit, and free cash flow, supported by operating leverage.

Still, the market reaction reflected what investors expected from the AI piece of the business. One report said Wall Street had hoped for AI semiconductor revenue around $16.36 billion for the quarter ending in early August, and that the guidance was read as “disappointing,” even as Broadcom maintained a fiscal 2027 revenue target of $100 billion for the AI-related business line. The same report also highlighted that, despite the drop, the stock still appeared expensive on common valuation measures, including a price-to-sales ratio near 24.9 and a price-to-earnings multiple above 60.

Beyond the numbers, the debate has shifted toward demand durability. The report pointed to a potential weakening announcement from AI software providers adjusting their pricing toward more consumption-based billing, arguing that enterprises may be rethinking usage when compute, electricity, and data-center costs climb. It cited examples of executives and customers voicing concerns about the practical cost of scaling AI workloads, which, if it spreads, could eventually translate into reduced purchases of semiconductors and networking equipment.

There is also a separate but related dependency in how quickly AI customers convert software demand into hardware orders. The report noted Alphabet’s TPU roadmap and described a purchase commitment by Anthropic for at least $21 billion worth of TPUs accessed through Broadcom, with deployment slated across 2026 and 2027, portraying it as evidence of ongoing procurement. But Broadcom itself, in its earnings release, did not break out any view specifically tied to changes in AI software pricing or customer cost complaints. Its filing language continued to emphasize general risks, including potential fluctuations in customer demand and the company’s ability to sustain margins.

What to watch next is whether the company’s AI semiconductor growth projections remain consistent when investors look past one quarter’s results. With the current fiscal Q3 forecast placing AI semiconductor revenue at $16.0 billion, the market will likely focus on whether Broadcom can defend its acceleration profile and whether any changes in customer budgeting show up in future guidance rather than only in third-party commentary. Any shift in AI procurement timing or mix could matter disproportionately for a stock that appears priced for sustained, high growth.