Broadcom Forecasts $230 Billion in AI Semiconductor Revenue in 2028. The Stock Could Reach $900 Per Share as a Result.

Broadcom expects AI semiconductor revenue to rise from $58 billion this year to $115 billion next year and approximately $230 billion in 2028, driven by ramping custom ASIC programs for hyperscalers including Alphabet, Meta, OpenAI, and Anthropic. The article estimates that, with roughly $58 billion of legacy-business revenue, Broadcom could reach $288 billion in 2028 sales and $144 billion in profit assuming a 50% margin. At a 30x earnings multiple, this scenario implies a $4.3 trillion market capitalization and a share price above $900, versus about $358 currently, though these projections depend on aggressive AI demand and margin assumptions.

Analysis

The relevant question is not whether custom accelerators grow, but whether AVGO can sustain economics through a hyperscaler procurement cycle. ASIC programs create unusually high switching costs after tape-out and software integration, supporting multi-year revenue visibility; however, they also concentrate pricing power in a handful of buyers that can dual-source designs or internalize more engineering over time. AVGO’s upside is therefore more sensitive to wafer/co-packaged-optics availability and customer deployment schedules than to aggregate AI demand.

Consensus likely underestimates the competitive impact on NVDA at inference and on lower-complexity training workloads. Every dollar redirected into TPU/other custom silicon reduces the addressable pool for merchant GPUs, but it also raises demand for networking, advanced packaging and HBM; the cleanest read-through beneficiaries beyond AVGO are TSMC and select optical-interconnect suppliers rather than GOOG or META, whose benefit arrives as lower depreciation per unit of compute. GOOG/META retain the option value of lower AI serving cost, but near-term capex intensity can still dominate reported FCF.

The article’s valuation logic is fragile: it assumes both a material margin step-up and a premium earnings multiple at a far larger revenue base. A more realistic downside case is that custom-chip mix expands while hyperscalers capture the cost savings through price negotiations, flattening gross-margin progression and compressing the terminal multiple. Over the next 1-3 months, the catalyst is management disclosure on funded backlog, supply commitments and customer-program ramp timing; over 6-18 months, watch whether custom-silicon deployments displace rather than supplement GPU clusters.

Contrarian framing: AVGO may be a superior AI exposure to NVDA only if custom silicon is truly incremental to total accelerator spend. If hyperscaler capex normalizes or a major program slips one production generation, AVGO’s concentrated revenue expectations leave less room for execution misses than the bullish narrative implies. The key falsifier is a reduction in AI semiconductor growth guidance, evidence of gross-margin dilution despite mix gains, or a material increase in customer concentration.

Key Decisions for Investors

• Maintain an AVGO overweight only on pullbacks or after verification of funded 12-18 month AI backlog; size against a 15-20% drawdown risk if one hyperscaler program slips. Add only if management confirms supply coverage and sequential AI revenue acceleration without gross-margin erosion.
• Express the custom-silicon share-gain thesis as long AVGO / short NVDA in modest notional for the next 6-12 months, rather than an outright AVGO chase. Exit the pair if NVDA demonstrates sustained inference share resilience or AVGO reports a program delay; the trade is invalid if ASIC demand proves additive rather than substitutive.
• Monitor GOOG and META capex guidance as a leading indicator, not a direct equity beneficiary. Raise AVGO exposure if either company signals accelerating internally deployed compute with stable total capex; reduce if spending shifts toward third-party cloud capacity or capex discipline.
• Avoid underwriting a multi-year AVGO target from peak-style margins until segment gross-margin disclosure, customer concentration and advanced-packaging capacity are independently confirmed. Set an alert for any AI guidance cut or margin miss at the next two earnings prints; either would warrant de-risking before the market revises terminal earnings assumptions.