Key Takeaways:
The AI supply chain's two most important companies are winning on opposite business models — and both are delivering numbers that would be extraordinary in any other cycle.
NVIDIA's 75% gross margins and TSMC's surging foundry revenue tell the same story from opposite ends of the AI supply chain: demand for AI compute is accelerating, not peaking.
"The buildout of AI factories, the largest infrastructure expansion in human history, is accelerating at extraordinary speed," Jensen Huang, chief executive officer of NVIDIA, said on the company's Q1 FY27 earnings call.
NVIDIA's Data Center segment reached $75.25 billion in the quarter, up 92% year over year, with networking alone at $14.8 billion — a 199% surge driven by InfiniBand, NVLink, and Spectrum-X. The company guided for $91 billion in Q2 FY27 revenue, excluding China entirely. TSMC, meanwhile, reported June revenue of NT$442.68 billion, up 67.9% from a year earlier, with first-half 2026 revenue climbing 35.6%. Advanced nodes at 7 nanometers and below accounted for 77% of TSMC's wafer revenue in Q4 2025, with 3nm at 28% and 5nm at 35%.
The two companies share a symbiotic relationship — NVIDIA's $119 billion in supply commitments flow directly into TSMC's order book — but their economics could not be more different. NVIDIA's fabless model delivers 75% non-GAAP gross margins, while TSMC's foundry business trades at a forward P/E of 27, roughly a 37% discount to NVIDIA's 43. China tensions remain the shared downside risk for both.
Designer Margins vs. Foundry Scale
NVIDIA collects rent on its CUDA ecosystem and platform lock-in, converting each GPU sale into a 75-cent profit on the dollar. That margin profile is unmatched in hardware and closer to enterprise software than chip manufacturing. TSMC's model is the inverse: lower margins per wafer but near-insurmountable scale advantages. The foundry's High Performance Computing segment alone generated NT$558.59 billion in Q3 2025, surpassing smartphones as its largest end market.
The gap in valuation multiples reflects this structural difference. NVIDIA trades at 43 times forward earnings, pricing in continued dominance of the AI accelerator market. TSMC at 27 times forward earnings — with shares up 79% over the past year — still carries a discount that suggests the market has not fully priced in the foundry's transition from cyclical commodity supplier to structural AI beneficiary.
The Shared Risk: China
Both companies face exposure to US-China export controls. NVIDIA has already excluded China from its Q2 FY27 guidance. TSMC serves Apple, AMD, and Broadcom alongside NVIDIA, but any escalation in semiconductor export restrictions would hit both the designer and the manufacturer. For TSMC, the risk is compounded by its concentration in Taiwan, though the company's geographic diversification efforts — including fabs in Arizona, Japan, and Germany — are underway.
Investor Takeaway
The AI trade is no longer a single-stock story. NVIDIA offers the margin leverage and the CUDA moat, but at a premium valuation that leaves little room for error. TSMC offers exposure to the same demand wave at a lower multiple, with the added buffer of a diversified customer base. For investors betting on AI infrastructure, the question is not which company wins — both are winning — but which business model fits their risk tolerance. NVIDIA's $91 billion guidance and TSMC's 67.9% June revenue growth suggest the answer, for now, is both.
This article is for informational purposes only and does not constitute investment advice.
