Broadcom is seeing a surge in demand tied to the AI boom, but investors are also noticing a key trade-off that could shape expectations for 2026. The company recently pointed to strong growth in AI-related orders, yet also acknowledged that gross margins on ASIC products are lower than margins on many of its non-AI offerings. That single detail was enough to stir uncertainty in the market, because booming AI revenue doesn’t automatically translate into equally booming profitability.
ASICs, or application-specific integrated circuits, are custom chips designed for a specific customer and workload. In today’s AI landscape, they’re increasingly used to power large-scale data center training and inference, often as alternatives or complements to general-purpose GPUs. As hyperscalers and major cloud providers push to control costs, energy use, and performance, custom silicon has become a major strategic priority, and companies like Broadcom are benefiting from that shift through increasing orders.
However, industry experts say lower margins on ASIC projects are not unusual, and the reasons are largely structural. Custom chips are typically built to a customer’s exact requirements, which gives buyers strong negotiating power on pricing. These deals can also involve heavier engineering and development work, longer production timelines, and tighter cost targets—especially when the customer is ordering at massive scale. As a result, the revenue opportunity can be substantial, but the margin profile often comes in below what a chipmaker might earn from more standardized products.
That’s why the market reaction matters: if more of Broadcom’s future AI growth comes from custom ASIC programs, the company could post impressive top-line results while seeing pressure on gross margin compared to a product mix tilted toward higher-margin, less customized offerings. In other words, AI-driven growth can still be very real, but the “quality” of that revenue—how much profit is generated per dollar—may differ depending on what exactly is being shipped.
The broader takeaway is that the AI chip race is no longer only about who sells the most, but also who can scale profitably. As ASIC orders rise across the industry, companies competing in custom silicon may need to balance volume growth with margin discipline, carefully managing engineering costs and negotiating terms that keep long-term profitability intact.
For investors and industry watchers looking ahead to 2026, the key question isn’t just whether AI-related demand continues—it’s how the product mix evolves, and whether higher volumes of custom AI silicon can offset the lower margin profile that often comes with customer-driven ASIC deals.
