EXCLUSIVE: Nvidia Keeps AI Profits, TSMC May Want a Bigger Cut

Nvidia reported a 71.1% fiscal-2026 gross margin, but MarketVector's Josh Kaplan argues that accelerating AI-chip demand could gradually shift a larger share of semiconductor profits toward foundries, particularly TSMC. Nvidia and other chip designers remain dependent on limited advanced manufacturing and packaging capacity, creating strategic leverage for TSMC and potentially Intel as volumes scale. The thesis is constructive for leading foundries, though it does not imply an immediate loss of Nvidia's pricing power or margins.

Analysis

The investable issue is not whether foundry pricing rises, but whether TSM can monetize scarcity faster than customers diversify designs or absorb higher wafer/package costs. A 100-200bp increase in TSM’s blended pricing on leading-edge wafers and advanced packaging would be materially accretive given its fixed-cost absorption, while the same move is modest for NVDA’s gross margin initially. The larger sensitivity is AMD: it lacks NVDA’s software-driven pricing umbrella and has less room to pass through higher manufacturing costs while competing for accelerator share.

Over the next 1-3 months, the key catalyst is evidence that advanced packaging—rather than wafer capacity—is constraining shipments. Watch TSM monthly revenue, CoWoS capacity-expansion commentary, lead times, and whether hyperscalers shift from buying standalone GPUs toward custom ASIC programs; the latter would expand AVGO’s addressable market but still reinforces TSM’s bargaining position. NVDA’s multiple is more vulnerable to any gross-margin guide below roughly 70% than to a modest absolute margin decline, since the market is underwriting both premium pricing and supply availability.

The contrarian view is that capacity additions can turn this from a scarcity rent into a volume-driven cycle. TSM’s customer concentration and geopolitical discount limit how much of the incremental economics can be capitalized into its multiple, while NVDA can offset foundry inflation through product mix, networking attach, and faster platform transitions. INTC is not a clean beneficiary until external foundry wins translate into credible advanced-node yields, utilization, and sustained positive foundry gross margin; capacity alone is not bargaining power.

Key Decisions for Investors

• Initiate a 6-12 month long TSM / short AMD pair: TSM captures both leading-edge wafer and packaging utilization, while AMD has greater margin and supply-allocation sensitivity. Target 15-20% pair return; exit if AMD demonstrates accelerator share gains without gross-margin dilution or TSM signals advanced-packaging supply normalization.
• Maintain NVDA core exposure but avoid adding into momentum until the next gross-margin guide: use a 3-month NVDA put spread, financed selectively with upside call overwrites, to protect against a sub-70% gross-margin narrative reset. The thesis is falsified if NVDA sustains margin guidance and backlog conversion despite higher supplier costs.
• Add AVGO on weakness rather than chase TSM as a second-order beneficiary of hyperscaler custom silicon. A shift toward ASICs diversifies AI compute demand away from merchant GPUs but does not remove advanced-node dependency; reassess if custom-chip revenue growth fails to accelerate over the next two earnings reports.
• Treat INTC as a watch item, not a foundry-scarcity long. Upgrade only upon independently verifiable external customer commitments, 18A yield milestones, and foundry utilization evidence; absent these, incremental capex risks depressing returns rather than creating pricing leverage.