Hock Tan Just Guided Broadcom's AI Revenue to Reach $230 Billion by 2028. Should You Believe Him?

Broadcom projects annual AI semiconductor revenue of $230 billion by fiscal 2028, versus an annualized run rate of just over $65 billion based on $16.7 billion in fiscal Q3 AI revenue. The company also lifted its 2027 AI-revenue outlook to $115 billion from a prior $60 billion-$90 billion range and says it has secured the supply chain needed to meet demand. The article estimates that hitting the AI target, alongside 10% non-AI growth and a 50% margin, could drive roughly $280 billion in 2028 revenue and support a $4.2 trillion valuation, about 150% above the current market cap.

Analysis

The investable issue is not whether custom silicon grows, but whether AVGO can convert design wins into a durable, high-margin platform rather than a lumpy handful-of-customers program. Hyperscaler ASIC adoption shifts value from merchant accelerators toward system-level integration: AVGO should capture compute plus high-speed networking content, while TSMC and advanced-packaging/HBM suppliers gain incremental utilization. The offset is customer concentration—any pause in GOOG, META, OpenAI, or Anthropic capex would create an outsized revenue and inventory-risk event relative to a diversified semiconductor model.

The market should discount management's long-range revenue target until customer purchase obligations, wafer allocations, and packaging capacity are independently visible. The article's implied valuation requires both unprecedented revenue scaling and sustained margin expansion despite a mix shift toward silicon with meaningful foundry, HBM, and advanced-packaging pass-through costs; revenue upside need not translate one-for-one to EPS upside. Over the next 1-3 months, quarterly AI revenue growth, backlog conversion, gross margin, and customer concentration disclosures matter more than the 2028 endpoint; 6-18 month confirmation comes from repeat generations of ASIC programs rather than first-generation ramps.

Consensus may be too binary on AVGO versus NVDA. ASIC share gains can pressure NVDA's inference economics, but NVDA retains software lock-in and may still benefit from aggregate AI infrastructure spend; the cleaner relative expression is AVGO versus a broad semiconductor basket, not an outright NVDA short. The key falsifier is a sequential deceleration in AVGO AI revenue accompanied by gross-margin compression or a reduction in hyperscaler capex guidance—evidence that supply availability has outrun end-demand.

Key Decisions for Investors

• Maintain/add AVGO only on post-earnings confirmation that AI revenue growth and consolidated gross margin both meet or exceed guidance; use a 6-12 month horizon. Size as a core long, but trim if management signals customer-led shipment phasing or gross margin declines materially for two consecutive quarters.
• Express custom-silicon supply-chain upside through a 6-12 month long AVGO / short SOXX pair, reducing broad semiconductor-beta exposure. Thesis requires AVGO to sustain AI growth while SOXX constituents face greater merchant-chip pricing competition; exit if AVGO underperforms SOXX by 10% following a clean earnings print.
• Watch TSM and MU as second-order beneficiaries, but do not initiate solely from the long-term forecast. Upgrade to longs only if company disclosures show incremental advanced-node/CoWoS utilization or HBM bit-demand revisions tied to ASIC ramps; these are the missing data needed to separate real volume from aspirational design pipelines.
• Avoid a directional NVDA short based solely on ASIC substitution. Consider it only if multiple hyperscalers explicitly shift inference spend from GPUs to internal ASICs while NVDA datacenter guidance decelerates; absent that evidence, NVDA's software and networking exposure can offset compute-share loss.