Broadcom Just Named Its Next Customer to Pass Google: Anthropic. Here's Why That Matters More Than the Earnings Beat.

Broadcom reported fiscal Q3 adjusted EPS of $3.32 and revenue of $29.59 billion, above consensus estimates of $3.24 and $29.36 billion, respectively, but shares fell after its $34.8 billion Q4 revenue outlook missed the $35.03 billion consensus. Management expects Anthropic to become its largest XPU customer in 2027, with Ironwood TPU compute purchases rising from 1GW in 2026 to 5GW in 2027; OpenAI's Jalapeno XPU deployment is projected to grow from 1.3GW in 2027 to 5GW in 2028. Broadcom projected adjusted EPS above $30 in fiscal 2028 versus the $25 analyst consensus, supported by accelerating custom AI-chip and networking demand.

Analysis

The key investable change is not a near-term revenue beat but validation that custom accelerators are becoming a second procurement lane for frontier-model developers rather than a single-customer Google program. That widens AVGO's addressable AI silicon pool and raises the durability of its ASIC engineering moat, but it also shifts the earnings debate toward execution: wafer allocation, advanced packaging capacity, yields, and the revenue captured per deployed megawatt. The disclosed deployment ambitions should not be treated as revenue until management clarifies shipment timing, chip content, and whether customer commitments are backed by funded data-center capacity.

For NVDA, custom ASIC adoption is a medium-term mix risk rather than an immediate volume cliff. Hyperscalers and model labs can deploy ASICs for steady-state inference/training workloads while retaining GPUs for frontier experimentation; the more immediate leakage is likely in incremental accelerator share and pricing power from 2027 onward. AVGO's networking franchise is an important hedge against that substitution, since scale-out AI clusters still require high-bandwidth switching regardless of compute architecture.

The post-results weakness suggests the market is discounting a familiar problem: AI upside is being capitalized before it converts into reported revenue. Over the next 1-3 months, AVGO needs either a raised AI revenue outlook, evidence that ASIC ramps are margin-accretive, or externally verified financing/capex plans from major model-lab customers to re-rate. The 6-18 month upside is meaningful if multiple customers qualify successive designs, but concentration risk rises if one lab delays its buildout, changes architecture, or faces funding constraints; a guide-down in AI semiconductor revenue or gross-margin compression would falsify the bullish thesis.

Key Decisions for Investors

• Accumulate AVGO on post-earnings weakness rather than chase a single-session recovery; target a 6-12 month holding period. Size against a stop/review trigger of a material reduction in AI semiconductor guidance or evidence that incremental ASIC revenue dilutes segment gross margin.
• Express the architecture-share theme as long AVGO / short NVDA in modest notional size over 12-18 months, rebalanced around earnings. The thesis is ASIC share gains in more predictable workloads; exit if NVDA demonstrates sustained acceleration in hyperscaler GPU orders or AVGO's customer ramps slip by more than one quarter.
• Do not underwrite projected megawatt deployments as contracted revenue until monitoring customer financing, data-center power availability, TSMC advanced-node allocation, and CoWoS/advanced-packaging lead times. Treat confirmation of funded capacity and production reservations as a catalyst to increase AVGO exposure.
• Maintain GOOG as a relative hedge rather than a direct beneficiary trade: broader custom-chip adoption can validate its internal TPU ecosystem, but external demand does not automatically accrue to Alphabet's P&L. Reassess only if Google discloses lower internal AI infrastructure cost or monetization improvement tied to TPU efficiency.