This AI Chipmaker Looks Like a Bargain With AI Revenue Set to Double Again Next Year

Broadcom expects AI semiconductor revenue to more than double in fiscal 2027 and double again in fiscal 2028, reinforcing substantial multiyear demand visibility for its custom AI chips. Fiscal Q2 revenue rose 48% year over year to $22.2 billion, above the company's prior $22 billion outlook, while net income increased 88% and AI semiconductor revenue more than doubled. Despite AI chips comprising more than 60% of sales and net margins exceeding 40%, Broadcom shares are up only 3% in 2026 versus a 13% S&P 500 gain, leaving the stock at roughly 19x forward earnings.

Analysis

The key underwriting question is not whether custom silicon demand grows, but whether Broadcom can convert a highly concentrated hyperscaler design-win pipeline into durable economics without giving back pricing in later generations. ASIC programs create multi-year revenue visibility once deployed, but customers retain meaningful negotiating leverage because volumes are concentrated and designs are bespoke; a single program delay can create an outsized quarterly miss. The relevant confirmation is disclosed backlog, customer prepayments, advanced-packaging capacity, and AI-semiconductor gross-margin progression—not management’s long-range revenue framing.

If custom accelerators gain share, AVGO is both a direct beneficiary and an indirect hedge on NVIDIA: hyperscalers deploying in-house compute still require high-radix Ethernet switching, DSP/optical connectivity and scale-out networking. This is incrementally constructive for ANET and selected optical suppliers such as COHR and LITE, while increasing the risk that NVDA’s accelerator revenue mix faces a slower growth rate beyond the next 12-24 months. The second-order constraint is packaging and networking deployment: weak availability of advanced packaging, optics, or power infrastructure would defer AVGO revenue rather than eliminate it.

Near term, the stock needs evidence that earnings estimates still understate the conversion of AI revenue into free cash flow after interest expense, integration costs and capital-intensity requirements. Over 1-3 months, quarterly guide cadence and customer-specific production ramps are the catalysts; over 6-18 months, the debate shifts to whether ASIC growth is additive to, or cannibalizes, merchant GPU spending. Consensus appears too focused on a headline valuation multiple: the appropriate multiple depends on whether AI margins expand while the acquired software base remains stable, rather than on semiconductor growth alone.

Key Decisions for Investors

• Accumulate AVGO on post-earnings volatility rather than chase pre-report strength; target a 6-12 month position sized against NVDA exposure. Thesis is validated by sustained AI-segment growth plus expanding consolidated free-cash-flow conversion; exit or reduce if the next two reports show AI growth deceleration with gross-margin erosion or a material cut to forward guide.
• Use a 6-12 month pair trade: long AVGO / short a calibrated NVDA notional, only after confirming AVGO’s next guide implies continued custom-silicon ramp. The trade captures incremental ASIC share gains while limiting broad AI-capex beta; main risk is NVIDIA retaining workload-performance superiority and AVGO suffering a hyperscaler program delay.
• Add a smaller basket of long ANET and COHR versus SOXX exposure for 6-18 months, as custom compute deployment raises Ethernet fabric and optical interconnect content even if accelerator vendor share shifts. Monitor hyperscaler capex guidance and optical-component lead times; a broad capex pause would invalidate the demand-throughput premise.
• Set an event alert around AVGO’s next earnings release for backlog, customer-concentration commentary, AI gross margin and free-cash-flow guidance. Do not add aggressively if management relies on multi-year targets without independently measurable production-volume, supply-capacity, or customer-deployment evidence.