Broadcom (AVGO) Q3 2026 Earnings Call Transcript

Broadcom reported fiscal Q3 revenue of $29.6B, up 86% year over year, with non-GAAP EPS rising 96% to $3.32 and free cash flow reaching $13.7B. AI semiconductor revenue surged 221% to $16.7B, prompting the company to raise fiscal 2026 AI revenue guidance to $58B from $56B; Q4 revenue guidance of $34.8B implies 93% growth and Q4 AI revenue of $21.7B implies 236% growth. Management outlined AI semiconductor revenue of roughly $115B in fiscal 2027 and $230B in fiscal 2028, supported by multiyear custom-chip programs with Google, Anthropic, OpenAI and Meta. Risks include data-center land, power and construction constraints, as well as substrate, HBM memory and laser supply bottlenecks; AI mix is expected to reduce Q4 gross margin to about 73% from 75% in Q3.

Analysis

The market should capitalize AVGO less as a cyclical semiconductor supplier and more as the control point for custom-compute ecosystems: design IP, packaging, switching, optics and increasingly capacity orchestration. That vertical stack raises switching costs for GOOG, META and frontier labs, while the Ethernet scale-up roadmap directly pressures NVDA's proprietary-networking premium and limits the addressable opportunity for standalone interconnect vendors. The less obvious beneficiary is optical-component capacity: industry scarcity should support pricing for COHR and LITE near term, although AVGO's internal laser expansion ultimately caps their upside.

The key earnings debate shifts from demand to conversion. The multi-year revenue path is credible only to the extent that customer sites receive power, permits, shells, HBM and financing on schedule; chips can be shipped but not recognized into the intended deployment cadence if those constraints bind. In addition, lower gross margin is not benign if memory inflation outruns operating leverage, and residual-value guarantees through third-party financing create a contingent-credit exposure that equity investors may begin valuing like off-balance-sheet leverage rather than pure semiconductor backlog.

Near-term, AVGO likely sustains estimate revisions through the next results cycle as supply additions and networking attach rates are modeled. Over 6-18 months, the larger risk is multiple compression if revenue becomes visibly dependent on two cash-burning labs or if data-center commissioning slips; management's customer claims should be verified against disclosed capex commitments, grid interconnects, and financing terms rather than treated as contracted revenue. Contrarian view: consensus may be underestimating the margin dilution and financing risk, but also underestimating Ethernet's ability to take share in GPU clusters, which broadens AVGO's upside beyond custom accelerators.

Key Decisions for Investors

• Maintain/add AVGO long on a post-earnings pullback rather than chase the opening gap; target a 6-12 month holding period. Thesis requires AI revenue conversion and operating-margin resilience, not merely bookings. Reduce if quarterly operating margin falls below management's mid-60s framework or if deployment timing causes a material guide-down.
• Initiate a 6-12 month pair: long AVGO / short NVDA in equal dollar beta-adjusted exposure. Custom ASIC adoption plus Ethernet scale-up should shift incremental AI-system economics toward AVGO; risk is NVDA retaining workload flexibility and software lock-in, or hyperscalers delaying proprietary silicon ramps.
• Establish a tactical basket long COHR and LITE for 1-3 months only if order commentary confirms persistent high-speed laser shortages. Treat AVGO's capacity ramp as the exit catalyst: internal supply expansion makes this a scarcity trade, not a durable structural long.
• Monitor AVGO credit disclosures and future XPV financing tranches before increasing gross exposure. Any material increase in residual-value guarantees, customer-payment support, or widening AVGO credit spreads would falsify the clean FCF narrative and justify trimming the core long.
• Avoid a standalone long in APO or BX solely on financing-platform headlines. Fee upside is plausible, but risk-adjusted returns depend on deal-level guarantees, asset ownership and investor protections that are not yet disclosed.