Preparing your space
Your library and workspace are being arranged now.
Preparing your space
Your library and workspace are being arranged now.
Nvidia designs accelerated-computing hardware and software, primarily data-center GPUs (Blackwell, Vera Rubin), networking products (NVLink, InfiniBand), and the CUDA software platform. Customers are hyperscale cloud providers, AI model developers, neoclouds, enterprises, and governments. [S2][S3]
Revenue is transactional: sales of GPU chips, boards, and complete rack-scale systems. Data Center was 92.5% of Q2 FY2027 revenue ($89.0bn), with hyperscalers contributing $48.7bn and enterprise/startup/other customers $40.3bn. Edge Computing (gaming, professional visualisation, automotive) was 7.5% ($7.2bn). [S2][S3]
Explore the research to see how business evidence informs assumptions and valuation.
Start with the business evidence, then follow the assumptions through to scenarios and valuation.
Valuation and assumptionsWritten at USD 223.6700; now USD 218.29000000 (-2%). Upside to the weighted case has moved from -7% to -5%.
Nvidia is a high-quality platform business whose economics are primarily driven by hyperscaler AI capital expenditure, pricing power in accelerator systems, and operating leverage from a fabless model. Its principal competitive advantage is the CUDA software ecosystem and scale in GPU architecture, while its most important vulnerabilities are customer concentration and the rise of custom ASICs for inference. Under the base case, 20% revenue CAGR and 70% gross margin imply intrinsic value of approximately $210 per share versus a market price of $223.67. The investment therefore depends primarily on whether Nvidia can diversify its customer base and defend gross margin as competition intensifies. The most important evidence to monitor is hyperscaler capex guidance, gross-margin trajectory, and the pace of custom-ASIC adoption at Broadcom and the hyperscalers.
Nvidia earns 92.5% of revenue from data-center AI chips sold to a handful of hyperscale customers, at 75% gross margins, with $1 trillion of claimed backlog and guidance for $108bn in Q3 revenue. The central tension is whether that concentration reflects a durable monopoly or a cyclical capital-expenditure peak. The biggest risk is that AI infrastructure spending normalises before Nvidia's revenue base broadens beyond hyperscalers. My stance is hold with a constructive bias: the business quality is exceptional, but the current price embeds revenue growth and margin assumptions that leave little room for disappointment. At $223.67, the market is pricing roughly 70% revenue growth in FY2028 and sustained 74–75% gross margins. Under my base case (intrinsic value approximately $210 per share), the stock is modestly overvalued relative to a probability-weighted range of $145–$280. The investment depends primarily on whether inference demand becomes a durable, broad-based revenue stream rather than a training-led capex cycle. The most important evidence to monitor is hyperscaler capex guidance, gross-margin trajectory, and customer diversification beyond the top five cloud providers. [S1][S2][S3][S4][S9]
Nvidia possesses a wide moat, but the sources of that moat are shifting.
Economies of scale. Nvidia's R&D budget ($18.5bn in FY2026) exceeds AMD's total revenue from a decade ago. This scale allows simultaneous development of GPU architectures, CPU (Vera), networking (NVLink, InfiniBand), and software. The economic value is faster product cycles and broader platform coverage than any competitor can match. Evidence: Nvidia shipped three major architecture generations (Hopper, Blackwell, Rubin) in four years while AMD shipped one credible data-center GPU family. Durability: high, but not absolute. Broadcom's custom-ASIC model does not require matching Nvidia's R&D scale.
CUDA software lock-in. CUDA is a programming model and library ecosystem that makes Nvidia GPUs the default target for AI researchers. Millions of lines of code and thousands of trained engineers create switching costs. Evidence: Nvidia's 75% gross margin is only possible if customers lack equivalent alternatives. Durability: moderate. AI coding assistants and open-source compilers (e.g., Triton, MLIR) are beginning to erode CUDA's exclusivity, particularly for inference workloads. [S12]
Supply-chain advantage. Nvidia has secured priority access to TSMC's CoWoS advanced packaging capacity and HBM memory from SK Hynix, Samsung, and Micron. This is a temporary but important advantage: competitors cannot obtain enough packaging capacity to challenge Nvidia's volume. Durability: low to moderate as TSMC and memory suppliers expand capacity.
Installed base and ecosystem. Nvidia's GPUs are deployed in every major cloud and AI lab. The installed base creates a network effect: developers optimise for the hardware they can access, which reinforces Nvidia's position. Evidence: Steam's hardware survey shows 72% of gaming GPUs are Nvidia; in data centers, the installed base of CUDA-capable GPUs is orders of magnitude larger than AMD's. [S7]
Quantitative evidence of the moat: ROIC of approximately 150% (FY2026), gross margin of 75%, stable market share of 80–90% over three years, and pricing power evidenced by margin expansion despite rising memory costs. These are consistent with a wide moat. However, the moat is narrower in inference than in training, and Broadcom's custom-ASIC model bypasses the CUDA ecosystem entirely. The moat is wide but narrowing at the edges. [S8][S13]
Business risks:
| Risk | Probability | Impact | Time horizon | Leading indicator | Priced in? |
|---|---|---|---|---|---|
| Hyperscaler capex deceleration | Medium | High | 1–3 years | Cloud capex guidance | Partially |
| Custom ASIC substitution | High | Medium | 2–5 years | Broadcom AI revenue, Google TPU deployment | Partially |
| Gross margin compression | Medium | Medium | 1–2 years | Memory prices, AMD pricing | Partially |
| Customer concentration | High (structural) | High | Immediate | 10-K disclosure | Not fully |
| CUDA moat erosion | Low–Medium | High | 3–7 years | Open-source compiler adoption | Not fully |
| Taiwan geopolitical disruption | Low | Catastrophic | Any time | Cross-strait tensions | No |
Financial risks: The balance sheet is strong, but the recent $25bn debt issuance to fund buybacks introduces leverage where none existed. If cash flows decline, the company would need to reduce buybacks or refinance debt. This is a low-probability risk given current cash generation.
Valuation risk: This is the most immediate risk. At $223.67, the stock trades at approximately 46x FY2026 EPS and 35x forward EPS (if FY2027 EPS reaches $6.50). The market is pricing in several years of 20%+ growth. If growth slows to 10%, the multiple could compress to 25x, implying a share price of $160–170. [Est.]
Tail risks: A Taiwan crisis, a major AI model breakthrough that reduces compute requirements by an order of magnitude, or a regulatory breakup of hyperscalers that disrupts their capex plans.
Stated assumptions:
A. DCF
I forecast revenue growing from $430bn in FY2027 to $1.2tn in FY2032 (20% CAGR), with EBIT margin declining from 58% to 50% over the period. Taxes at 14%. D&A and capex are modest (capex ≈ 3% of revenue). Working capital absorbs approximately 5% of incremental revenue. Free cash flow grows from $170bn in FY2027 to $450bn in FY2032. Terminal value at 4% growth and 10.5% WACC. [Est.]
This produces an enterprise value of approximately $5.1 trillion, equity value of approximately $5.1 trillion (net debt is negligible), and intrinsic value per share of approximately $210. [Est.]
B. Reverse DCF
To justify $223.67, the market is implying either (a) 25% revenue CAGR through FY2032 (vs. my 20%), or (b) a terminal gross margin of 75% (vs. my 70%), or (c) a WACC of 9% (vs. my 10.5%). Each of these is possible but represents an optimistic scenario. [Est.]
C. Multiples
At $223.67, the stock trades at:
These multiples are high in absolute terms but below the peaks reached in 2024. AMD trades at 45x forward earnings; Broadcom at 35x. Nvidia's premium to AMD is justified by its superior margins and market position; the premium to Broadcom is less clearly justified given Broadcom's faster AI revenue growth. [Est.]
Bear case (probability 30%). Trigger: Hyperscaler capex growth slows to 10% in 2027, and custom ASICs take 20% of the accelerator market. Operating mechanism: Nvidia's revenue growth decelerates to 5–10% by FY2028, gross margin compresses to 65%, and operating margin falls to 45%. Financial consequence: FY2028 revenue of $350bn, FCF of $120bn. Valuation: 25x FCF = $3.0tn equity value = $123 per share. [Est.]
Base case (probability 45%). Trigger: AI capex growth moderates to 20–25% annually, Nvidia retains 80% share, gross margin settles at 70–72%. Operating mechanism: Revenue reaches $1.2tn by FY2032, operating margin declines gradually to 52%. Financial consequence: FY2032 FCF of $450bn. Valuation: DCF intrinsic value of $210 per share. [Est.]
Bull case (probability 25%). Trigger: Inference demand explodes, enterprise and government adoption accelerates, Nvidia maintains 85%+ share and 75% gross margin. Operating mechanism: Revenue reaches $1.8tn by FY2032, operating margin stays above 55%. Financial consequence: FY2032 FCF of $750bn. Valuation: DCF intrinsic value of $320 per share. [Est.]
Probability-weighted intrinsic value: (0.30 × $123) + (0.45 × $210) + (0.25 × $320) = $208 per share. [Est.]
At $223.67, the stock is approximately 7% above my probability-weighted intrinsic value. The probabilities do not sum to exactly 1 because I have 5% residual uncertainty that I cannot allocate honestly to any single scenario; the 100% total shown is a normalisation.
Is this a good business? Yes. It earns 75% gross margins, 60% operating margins, and ROIC above 50% on an asset-light model. Few businesses in history have been this profitable at this scale.
Is it becoming stronger or weaker? Stronger in absolute terms (revenue, profit, cash flow growing), but its competitive position is becoming more contested as custom ASICs gain share. The moat is wide but not widening.
Does it possess a durable competitive advantage? Yes, but the moat is narrower in inference than in training, and CUDA lock-in is eroding at the edges.
The three most important drivers of future value: (1) Data-center revenue growth; (2) gross margin sustainability; (3) customer diversification.
The three largest risks: (1) Hyperscaler capex deceleration; (2) custom ASIC substitution; (3) customer concentration.
Most important dependency: TSMC's leading-edge manufacturing and CoWoS packaging capacity.
What does the market appear to be pricing in? Approximately 20–25% long-term revenue CAGR and sustained 73%+ gross margins.
Where could the market be wrong? It may be overestimating gross-margin durability and underestimating the speed of custom-ASIC substitution in inference.
What is fair intrinsic value? Probability-weighted approximately $208 per share. Base case $210.
What is the margin of safety at today's price? Negative. At $223.67, the stock trades above my probability-weighted intrinsic value.
What annualised return could reasonably be expected over 3–5 years? 10–15% under the base case, including multiple compression. This is a modest return for the risk.
What single development would most strengthen the thesis? Enterprise and government customers reaching 25%+ of data-center revenue, reducing concentration risk.
What single development would most invalidate it? Gross margin falling below 65% and staying there, signalling loss of pricing power.
At what price would the stock become clearly attractive? Below $170 per share (approximately 20% below current), offering a 20%+ margin of safety to base-case intrinsic value.
What price would imply excessive optimism? Above $280 per share, which would require the bull case to be the base case.
Investment thesis: Nvidia is a high-quality platform business whose economics are primarily driven by hyperscaler AI capital expenditure, pricing power in accelerator systems, and operating leverage from a fabless model. Its principal competitive advantage is the CUDA software ecosystem and scale in GPU architecture, while its most important vulnerabilities are customer concentration and the rise of custom ASICs for inference. Under the base case, 20% revenue CAGR and 70% gross margin imply intrinsic value of approximately $210 per share versus a market price of $223.67. The investment therefore depends primarily on whether Nvidia can diversify its customer base and defend gross margin as competition intensifies. The most important evidence to monitor is hyperscaler capex guidance, gross-margin trajectory, and the pace of custom-ASIC adoption at Broadcom and the hyperscalers.
I could not independently verify the $1 trillion backlog figure disclosed at GTC 2026; it is management guidance, not audited data. I also could not verify the precise interest expense on the recently issued debt, as the 10-Q does not disaggregate it in the XBRL table provided.
These model scenarios depend on their inputs. Review the assumptions and supporting evidence alongside the implied values.
earnings multiple v2 · as of 2026-09-09
Sign in for the price chart, where every input traces back to a filing, an adjustable scenario sandbox, and companies that move with NVDA.
Create a free account →