Broadcom CEO Hock Tan Says Demand Already Exceeds Broadcom's $115 Billion AI Forecast. Here's What That Means for the Stock.

Broadcom CEO Hock Tan said AI revenue is expected to double to approximately $115 billion in fiscal 2027 and double again to roughly $230 billion in fiscal 2028, with demand already exceeding the 2027 supply outlook. The company reported fiscal Q3 sales growth of 86% to $29.6 billion and non-GAAP EPS growth of 96% to $3.32, driven by custom AI processors. Management expects fiscal 2028 EPS to exceed $30, more than six times fiscal 2025 earnings, supported by ASIC demand from customers including Alphabet, Anthropic, and OpenAI.

Analysis

The investable issue is not simply ASIC demand, but whether AVGO can convert booked design wins into durable economic profit. Custom silicon creates high switching costs once a model, compiler stack, packaging design, and data-center network are qualified, supporting multi-year revenue visibility; however, hyperscale customers will use their concentrated purchasing power to capture part of the cost-per-token savings. The key debate over the next 1-3 months is therefore gross-margin mix and customer prepayment/working-capital terms, not top-line demand alone.

AVGO's custom-chip scale is a selective negative for NVDA rather than a wholesale displacement thesis. ASICs are most effective for stable, high-volume internal workloads, while NVDA retains the advantage in rapidly changing frontier-model architectures, broad software compatibility, and external cloud rental markets; the second-order loser is likely merchant accelerator demand at the largest hyperscalers, not GPU demand overall. GOOG benefits if internally designed compute lowers Gemini inference costs faster than peers, but capex intensity could obscure that benefit in reported free cash flow for several quarters.

Consensus appears to be treating the long-range outlook as incremental certainty even though the path depends on advanced packaging, HBM availability, customer data-center power delivery, and a small number of counterparties maintaining extraordinary capex. A shortfall in any one component can defer revenue rather than destroy it, but deferrals would matter materially to a stock priced on forward earnings duration. For 6-18 months, monitor whether AI semiconductor revenue grows faster than total operating cash flow: divergence would signal customer-financing, inventory, or lower-margin mix risk.

The contrarian view is that ASIC adoption may expand the total AI-compute budget rather than cannibalize it: lower unit economics can unlock inference workloads that are uneconomic on general-purpose accelerators. That supports a barbell of AVGO and NVDA rather than an outright substitution trade, while favoring network and optical beneficiaries such as ANET and LITE if compute clusters continue to scale. The thesis is falsified by a material cut to hyperscaler capex guidance, evidence of HBM/packaging constraints extending beyond one or two quarters, or AVGO AI gross margin falling despite revenue growth.

Key Decisions for Investors

• Initiate a measured 6-12 month long AVGO position only after verifying the current valuation against fiscal-2028 EPS expectations; use a 25-30% downside risk budget versus a base case of sustained AI revenue conversion and multiple stability. Do not chase solely on long-dated management targets—add on evidence of margin resilience and firm customer purchase commitments.
• Express the ASIC-share-gain thesis as long AVGO / short a small basket of merchant AI-compute exposure, weighted toward NVDA only if NVDA's hyperscaler GPU revenue or forward supply commentary weakens. Keep the hedge ratio low: ASIC deployment can be incremental to GPU demand, making a large AVGO-NVDA relative-value position vulnerable to a broad AI-capex upside surprise.
• Add GOOG on 6-18 month weakness if disclosure demonstrates improving AI inference cost per query without a corresponding acceleration in capex; the catalyst is margin leverage from TPU utilization rather than chip sales. Falsify on rising depreciation/capex with no improvement in cloud operating margin or AI monetization metrics.
• Set an earnings watch item for AVGO: require AI segment growth, consolidated gross margin, inventory days, customer-concentration commentary, and operating cash-flow conversion to validate the thesis. A revenue beat accompanied by weaker gross margin or working-capital outflow should be treated as a de-risking signal rather than a reason to add.