NVIDIA (NVDA) Stock Analysis & AI Equity Report

The core valuation tension is duration and cycle positioning, not current cash generation. The market has allocated 99.8% of enterprise value to Compute & Networking, which generates 89.6% of group net sales (USD 193.5 billion) and USD 130.1 billion of segment EBIT at a 67.3% operating margin. The stock reflects a 34.99x trailing 2026 EV/EBIT multiple but only 14.7x on 2027e EBIT of USD 240,008m — a meaningful discount to the peer median of 20.8x — and consensus EBIT must jump 84.0% from USD 130,387m in FY26 to USD 240,008m in FY27e, almost entirely sourced from Compute & Networking. That bridge requires hyperscaler capex to hold near USD 725bn in 2026, up 77% year on year. On a reverse-valuation basis, a conservative 20x mature technology framework applied to current consolidated EBIT of USD 130.4 billion supports approximately USD 2.6 trillion of enterprise value, roughly 57% of the current allocation; the remaining 43% depends entirely on the forward consensus that EBIT reaches USD 240 billion by FY2027. Our 24.0x fair EV/EBIT multiple on 2027e EBIT of USD 240,008m (237.0 USD, 48% weight), blended with a 28.0x P/E on 2027e EPS of 8.65 (242.2 USD, 32% weight) and the ROE-versus-P/BV scatter framework (378.8 USD, 20% weight), yields the 267.0 USD weighted target. The scenario base case implies USD 267.30 per share (+84.7%), the bear case compresses EBITDA to USD 150,000m at 15.0x for USD 92.54 (-36.0%), and the bull case reaches USD 343.93 (+137.7%).

NVIDIA currently trades at 14.7x 2027e EV/EBIT, a meaningful discount to the peer median of 20.8x, and our selected fair multiple of 24.0x underpins most of the re-rating case for the target price.

Scaling pool at the core of the enterprise: Compute & Networking represents 89.6% of group net sales (USD 193.5 billion) and 99.8% of positive segment operating profit (USD 130.1 billion) at a 67.3% operating margin, and under the fallback EBIT allocation framework is assigned 99.8% of group enterprise value, or approximately USD 4.55 trillion. The market values the segment because it captures a large share of the global AI infrastructure buildout — a highly profitable, immediate cash-generating business rather than a speculative option — with investors assigning a mid-30x trailing EBIT multiple because they expect the shift from AI training workloads to AI inference workloads to keep expanding the total addressable market. The moat rests mainly on system-level integration rather than pure silicon performance: the CUDA software ecosystem provides a strong developer-level moat, while the physical moat in 2026 relies on proprietary networking — NVLink and InfiniBand — with systems like the GB300 NVL72 functioning essentially as a single, large logical GPU. Maintaining these margins requires controlling the bill of materials, particularly HBM3e and HBM4 memory sourced from SK Hynix, Micron and Samsung, and TSMC's advanced CoWoS packaging costs.

All figures in USD millions unless noted; per-share data in USD.