DRIVERS

REVENUE
- Stage 1 (Y1–5): 25.4% flat-rate equivalent (modeled as such in Specula). In practice, a declining path is more defensible: Year 1 ~35%, tapering to ~17% by Year 5. The blended average of 25.4% reflects Blackwell Ultra ramp continuing through FY2027, Rubin architecture production shipments commencing H2 FY2027, inference deployment scaling faster than training, and sovereign AI demand (Middle East, India, Europe) adding incremental volume outside the hyperscaler channel. The China headwind (~$25B in FY2025 vs. negligible in FY2026) is treated as a permanent structural impairment and embedded in the base.
- Stage 2 (Y6–10): 10%. Normalization as AI infrastructure matures. Inference expansion, DRIVE automotive ramp, and NVIDIA AI Enterprise software partially offset the deceleration in data center hardware. Hyperscaler custom silicon (Google TPU, Amazon Trainium, Meta MTIA) assumed to capture 15–20% of workloads previously on NVIDIA hardware within the hyperscaler channel. CUDA stickiness holds outside hyperscalers.
- Terminal: 3.5%. Above nominal GDP. Justified by NVIDIA's embedded position in AI inference infrastructure — a long-duration, recurring demand base analogous to cloud infrastructure. Not above 4% — that would require compounding faster than the global economy indefinitely.

OPERATING MARGIN
- 60% in Years 1–5, declining to 55% by Year 10, holding at 55% terminal. FY2026 operating margin was 60.4% (GAAP), including the $4.5B H20 charge. Normalized, it would have been approximately 62%. The path to 55% reflects: (a) gross margin compression as Blackwell/Rubin rack-scale systems carry higher system-level cost than prior HGX configurations, (b) R&D intensity increasing in absolute terms as NVIDIA funds Rubin, Feynman, and the software stack, and (c) modest pricing pressure as inference competition intensifies. SBC of $6.4B in FY2026 is already expensed in GAAP operating income — no adjustment required.

REINVESTMENT / SALES-TO-CAPITAL
- Sales-to-capital ratio: 2.0x (constant). NVIDIA's fabless model means tangible capex is structurally low ($6.0B on $216B revenue in FY2026, or 2.8%). True reinvestment is R&D-driven. Post-capitalization, the invested capital base grows from $38B (FY2022) to $193B (FY2026). A sales-to-capital ratio of 2.0x is conservative relative to recent history but appropriate as a long-run assumption given the increase in capital complexity (rack-scale systems, US domestic manufacturing investments, Groq IP acquisition at $13B).

ROIC vs WACC
- ROIC path: 50% in Years 1–5 (post-R&D capitalization), declining to 35% by Year 6–10, stabilizing at 25% terminal. The current realized ROIC (post-cap) is substantially above these levels — the decline path embeds the assumption that pricing power normalizes as competition intensifies and the invested capital base expands. At 25% terminal ROIC against a WACC of 9.3%, the spread remains meaningfully positive, consistent with a durable but competed moat.
- WACC: 9.31% (Specula computed). Decomposition: risk-free rate 4.04% (10Y UST), unlevered beta 1.27, levered beta 1.316 (at 4% target debt ratio), ERP 4.17% (Damodaran implied), no country risk premium, cost of equity 9.53%, cost of debt 4.79% (RF + 75bps spread), marginal tax rate 13%, WACC = 9.31%. This is a disciplined, market-implied estimate — not inflated by historical equity risk premiums.

TAX
- 13% marginal rate (constant). FY2026 effective rate was 15.1%, up from 13.3% in FY2025, reflecting lower relative SBC deductions as income scales. The 13% assumption is slightly below recent effective rates but defensible: FDDEI deductions, Israeli reduced statutory rate on qualifying income (a meaningful benefit given ~6,000 employees in Israel and Mellanox's integration), and R&D tax credits are structural, not one-time. The One Big Beautiful Bill Act (OBBBA), enacted July 2025, modifies certain provisions — the net effect has been recognized in FY2026 results. Upside risk: if Pillar Two minimum tax implementation accelerates, the effective rate could drift toward 17–18%, reducing intrinsic value by approximately 8–10% under base assumptions.

DEBT POLICY
- Target debt ratio: 4% (debt as percentage of total capital). NVIDIA is a net cash business with $62.6B in cash and marketable securities against $8.5B in notes outstanding as of January 25, 2026 — a net cash position of approximately $54B. The 4% debt target reflects minimal structural leverage. NVIDIA maintains fixed-rate senior notes ($8.5B outstanding, maturities from 2026 to 2060) for capital structure efficiency, not necessity. A $25B commercial paper program was established in January 2026 but carries no outstanding balance. No financial distress risk on any horizon.
- Cost of debt spread: 75bps over risk-free (total pre-tax cost ~4.79%). Consistent with investment-grade pricing (Aa1/AA equivalent).