NVIDIA Corp. From hardware monopoly to platform utility — the toll-keeper of global AI compute

01 The Investment Thesis

The core valuation thesis rests on NVIDIA's evolution from a hardware monopoly into a structural toll-keeper of global AI computing. The near-term projections capture an unmatched scaling phase, pushing revenues above the $1.1 trillion mark by 2036. While a trillion-dollar top-line is historically unprecedented, accounting for structural monetary expansion (M2) and long-term inflation over a ten-year horizon — added to NVIDIA's strategic importance in the AI race — brings this figure into a plausible macroeconomic reality.

To fuel this growth, the reinvestment profile follows a rational bell-curve trajectory. Net reinvestment peaks rapidly at roughly $52 billion by Year 2 to secure supply-chain constraints, foundry capacity and advanced-packaging allocations, with a highly efficient sales-to-capital ratio of 2.0 through the transition years. As the industry matures, return on capital step-downs from an unsustainable 63.6% to a terminal 20.0%. That 20.0% steady-state ROIC is the numerical anchor of the moat: it acknowledges that while hardware margins face pressure from lower-cost custom ASICs and tech-buyer pushback, the entrenched CUDA software layer prevents absolute margin destruction, ensuring structural value creation well above the 8.00% terminal cost of capital.

The cash-flow engine is shaped by the convergence of operational and fiscal levers. Operating margins — excluding R&D as an operating expense and adding it to reinvestment — are modeled to decay from 70.0% down to a mature 40.0%, a compression that is mathematically inevitable as high-margin AI training saturates and gives way to competitive AI inference. Simultaneously, the model captures fiscal reality by scaling the effective tax rate from an optimized 16.0% up to a 24.0% domestic and global marginal rate, acknowledging the dilution of R&D tax credits against an immense pre-tax income base.

02 The Valuation Model

All figures in US$ millions except per-share and percentage data. FCFF stays positive throughout — operating margin compresses from 70% to 40% as AI training saturates, while reinvestment follows a bell-curve that peaks early to secure supply.

Market price at valuation date: $207.00. Implied upside to intrinsic value: approximately 1%.

03 The Sensitive Assumptions

• Terminal Margin A 40% operating margin sustained to perpetuity — down from 70% today — is the central modeled call. The compression is inevitable as AI training saturates and inference competition intensifies; the floor is defensible only on the strength of the CUDA moat.
• Terminal ROIC The 20.0% steady-state return on capital is the numerical anchor of the moat, held ~12pp above the 8% cost of capital. Lower-cost custom ASICs and tech-buyer pushback are the structural threats to this excess return.
• Reinvestment Path The bell-curve peaking near $52bn by Year 2, governed by an efficient 2.0 sales-to-capital ratio, assumes capacity is built without disruption. Foundry, supply-chain or advanced-packaging bottlenecks would defer FCFF and compress value.
• Hyperscaler CapEx Because the market already prices a historic transition to a structural platform utility, the margin of safety is exceptionally thin — leaving the stock highly sensitive to operational bottlenecks or deviations in the near-term capital spending of its core hyperscaler clients.