Broadcom vs. Nvidia: 3 Key Metrics Point to the Stronger AI Chipmaker to Buy After Earnings

Nvidia reported 106% year-over-year adjusted revenue growth and 120% diluted EPS growth, while guiding current-quarter revenue to $108 billion, about $2.5 billion above consensus. Broadcom delivered 86% adjusted revenue growth and 96% adjusted EPS growth; its $34.8 billion near-term revenue outlook was slightly below the $35.03 billion consensus, though management projected AI semiconductor revenue could reach $115 billion in 2027 and $230 billion in 2028. Nvidia is presented as the more attractive AI-chip investment due to its lower forward P/E, $89 billion in quarterly AI-chip revenue versus Broadcom's nearly $21 billion, and projected 70% fiscal-2028 revenue growth, despite expected gross-margin compression to 71%-72%.

Analysis

The near-term earnings-revision setup favors NVDA: a current-quarter guide above consensus is immediately monetizable, while AVGO's incremental upside rests more heavily on a multi-year custom-silicon ramp that analysts cannot yet underwrite at full value. The valuation discount in NVDA is therefore not simply a growth-quality signal; it reflects investor concern that accelerator mix, memory inflation, and large-customer concentration eventually normalize its exceptional profitability. Over the next 1-3 months, estimate revisions and hyperscaler capex commentary should matter more than distant revenue targets.

Memory-cost inflation is a non-obvious beneficiary channel. NVDA's projected margin trough implies that HBM and advanced-memory suppliers, particularly MU, can capture a disproportionate share of AI-system economics even if GPU unit demand remains robust; AVGO's custom ASIC mix is similarly memory intensive. Conversely, sustained gross-margin pressure would undermine the premise that semiconductor AI revenue deserves software-like multiples, creating downside beta for both names and especially for richly valued suppliers with unproven 2027-28 demand visibility.

The structural competitive issue is not NVDA versus AVGO headline growth, but whether custom ASIC deployments become a material substitute for merchant GPUs at the largest cloud customers. AVGO wins if customers shift stable, high-volume inference workloads to lower-cost proprietary silicon; NVDA retains the premium if model architectures, software lock-in, and faster deployment cycles keep workload requirements changing too quickly for ASIC design cycles. The article's long-range projections should be treated as management scenarios until customer-specific volume commitments, foundry capacity, and associated networking attach rates are independently visible.

Contrarian view: the market may be underpricing a near-term AVGO reset if investors begin discounting its long-dated AI opportunity less aggressively after a below-consensus quarter, while it may be overpricing NVDA's ability to pass through memory costs without any multiple penalty. The key falsifier for the NVDA relative thesis is two consecutive quarters of gross margin below the stated trough range or evidence that cloud capex is reallocating from GPUs to internally designed ASICs rather than expanding overall.

Key Decisions for Investors

• Initiate a 1-3 month relative-value position: long NVDA / short AVGO in beta-adjusted dollar amounts after confirming post-earnings estimate revisions remain positive for NVDA and flat-to-negative for AVGO. Target 8-12% relative outperformance; stop if AVGO's next-quarter revenue guide moves above consensus by more than 3% or NVDA gross-margin guidance falls below 71%.
• Add MU to the AI basket rather than increasing outright NVDA exposure: accumulate over the next 1-2 months on evidence of improving HBM pricing and supply tightness. The thesis is that memory-content inflation transfers margin from AI-chip designers to memory vendors; exit if HBM contract pricing fails to rise or NVDA/AVGO indicate successful cost offsets.
• Use a defined-risk hedge on broad AI-semiconductor exposure via SOXX puts 3-6 months out, sized against long NVDA/MU exposure. The hedge becomes valuable if hyperscaler capex guidance weakens or the market rerates AI hardware on gross-margin compression rather than revenue growth.
• Do not underwrite AVGO's long-dated AI scenario as a standalone long until customer concentration and custom-chip volume commitments are disclosed. Set an alert for confirmed large hyperscaler ASIC wins, associated networking revenue acceleration, and foundry capacity reservations; those data would invalidate the short leg of the NVDA/AVGO pair.