선택적 제품 분석
허용하는 경우 개인정보 보호를 고려한 제한적 분석을 사용해 어떤 리서치 여정이 효과적인지 파악합니다. 광고 추적기나 세션 녹화는 사용하지 않습니다. 환경설정에서 이 선택을 변경할 수 있습니다. 자세한 내용은 개인정보 처리방침.
Accelerating transition to continuous 24/7 background agentic AI workloads and expansion of neocloud cluster capacity from 3 GW to 8 GW, driving compounding demand for high-margin NVL72 and Vera Rubin rack-scale systems.
We rate NVIDIA Corporation a Strong Buy with a Fair value of $374.88 per share, representing 68.7% upside from the report-date price of $222.27, underwritten by durable platform lock-in across accelerated computing infrastructure.4
Our rating is Strong Buy. For an ordinary investor, this indicates that NVIDIA represents a compelling opportunity where equity markets significantly undervalue the company's multi-year cash flow generation and durable platform lock-in.
As of September 18, 2026, NVIDIA trades at $222.27 per share. Against this level, our house conclusion identifies a Fair value of $374.88 (representing 68.7% upside) and a Potential value of $586.75 (representing 164.0% upside).4
The primary risk that would alter our investment thesis is a synchronized capital expenditure freeze by top cloud hyperscalers, who account for 36% of FY2026 revenue. If hyperscalers pause cluster buildouts to digest existing infrastructure while shifting internal workloads to custom silicon, revenue growth would decelerate below 20% and gross margins would fall below 70%, impairing operating profitability.56
The strongest reason the thesis succeeds is NVIDIA's architectural dominance in complete rack-scale computing systems, fortified by the proprietary CUDA software platform. Turnkey delivery of NVL72 and Vera Rubin platforms enables customers to achieve superior token cost economics, expanding projected revenue to $680.18B in FY2028 with operating margins exceeding 61%, directly compounding cash flow per share toward our Fair value.5
CUDA's two-decade developer ecosystem and deep library integration make NVIDIA's architecture the universal standard for AI model training and agentic inference, creating high switching costs that protect operating margins above 58%.5
Advance capacity reservations with TSMC and packaging partners secure guided ~70% growth in FY2028, converting massive backlog into cash flow across the Vera Rubin ramp.5
Turnkey NVL72 and Vera Rubin liquid-cooled rack platforms maximize compute density per gigawatt, establishing unmatched total cost of ownership leadership that standalone accelerator chips cannot match.5
Strong Buy: NVIDIA's proprietary CUDA ecosystem, NVLink interconnect architecture, and secured supply chain capacity protect premium margins and multi-year cash flow compounding across global AI infrastructure deployments.5
NVIDIA pioneered accelerated computing to help solve the most challenging computational problems. NVIDIA is now a data center scale AI infrastructure company reshaping all industries. NVIDIA Corporation specializes in markets where its computing and AI infrastructure platforms can provide tremendous acceleration for applications. These platforms incorporate processors, interconnects, software, algorithms, systems, and services to deliver unique value. Its platforms address four large markets where its expertise is critical: Data Center, Gaming, Professional Visualization, and Automotive.
The Compute & Networking segment includes its Data Center accelerated computing and networking platforms and AI solutions and software, and Automotive platforms and autonomous and electric vehicle solutions including software. The Graphics segment includes GeForce GPUs for gaming and PCs, and Quadro/NVIDIA RTX GPUs for enterprise workstation graphics.
A severe stress scenario occurs if Tier-1 cloud hyperscalers conclude their initial cluster deployments and simultaneously pause infrastructure capex to digest compute capacity while ramping captive internal silicon (e.g. Jalapeno). Because Customer One (22.0%) and Customer Two (14.0%) account for 36.0% of consolidated revenue, such a pause would decelerate Compute and Networking growth below 20%. Simultaneously, yield friction on multi-die 2.5D/3D packaging and NVL72 liquid-cooled rack integration would depress gross margins below 70%, compressing consolidated operating margins toward 43%. If sustained systematically high volatility expands the cost of equity toward 14.67% and de-rates the terminal multiple toward peer levels of 18.5x (AVGO at 18.45x), cash flow generation would compress to the downside valuation boundary of $257.82.
In the Bear scenario, datacenter expansion pauses post-initial cluster buildouts as top hyperscalers (representing 36.0% of FY2026 revenue) digest hardware capacity and ramp internal custom silicon (e.g. Jalapeno). Growth decelerates below 20% in the medium term, and assembly yield friction in complex NVL72 rack systems compresses gross margins below 70%. Consolidated operating margins contract toward 43%, and valuation multiples compress toward peer levels, establishing the downside boundary at $257.82.
Customer concentration of 36.0% across top two customers, leads to severe top-line deceleration if datacenter expansion pauses post-initial cluster buildouts, missing guided ~70% growth.56
Management guides Q4 FY2027 gross margin down to 71%-72% and FY2028 to 72%-73%. Yield friction in complex NVL72 rack-scale systems and packaging constraints dropping gross margin below 70% reduces consolidated operating margin by >6pp.5
NVIDIA trades at 24.15x forward P/E whereas peers trade at lower multiples (AVGO at 18.45x). De-rating from the assumed 30.0x terminal multiple toward peer averages contracts forward earnings fair value by 20.0%.34
With raw beta of 2.2478 and moderate Ke of 12.6706%, a 200 bps increase in discount rates de-rates long-duration cash flows, impacting discounted earnings and FCFF DCF by approximately 20%.2
투자 논리를 약화시킬 수 있는 다섯 가지 변화입니다.
Customer concentration surpassing 45% would signal rising vulnerability to individual hyperscaler capex digestion, weakening enterprise and sovereign diversification.
현재:36.0% combined in FY2026 (Customer One 22.0%, Customer Two 14.0%)
다음이면 경고>45.0% combined share
Gross margins breaking below 70% would reflect acute packaging yield friction in NVL72 systems and pricing concessions against custom internal silicon.
현재:74% ±50 bps guided for Q3 FY2027
다음이면 경고<70.0%
A collapse in compute growth below 20% would demonstrate an abrupt pause in datacenter infrastructure expansion, falsifying forward earnings trajectories.
현재:~70% guided for FY2028 and $193.48B in FY2026
다음이면 경고<20.0% YoY
Multiple contraction toward broad semiconductor peer averages would indicate structural market de-rating and terminate premium hyper-growth compounding expectations.
현재:24.15x forward P/E as of 2026-09-19
다음이면 경고<18.5x (semiconductor peer level, AVGO at 18.45x)
A precipitous drop in China revenue below 3% driven by tightened export restrictions would eliminate international operating leverage if not offset by domestic expansion.
현재:$19.68B in FY2026 (9.1% of consolidated revenue)
다음이면 경고<3.0% of revenue
NVIDIA possesses a wide economic moat rooted in its full-stack accelerated computing architecture, combining specialized silicon, the proprietary CUDA programming model, and ultra-high-bandwidth NVLink interconnects. This ecosystem creates severe switching barriers for software developers and enterprise customers. Over two decades of library development, runtime optimization, and developer mindshare across frameworks such as cuDNN and TensorRT mean that algorithms and AI applications are natively optimized for NVIDIA hardware. Even frontier labs investing in captive internal silicon continue scaling primary frontier models on NVIDIA infrastructure.
The platform moat is further entrenched by rack-scale system engineering. The transition from individual discrete accelerator cards to third-generation NVL72 rack-scale systems creates unified computing fabrics capable of seamlessly shifting between data processing, pre-training, post-training, and continuous agentic inference. Competing silicon solutions from merchant peers and captive in-house accelerators face substantial integration friction because they lack the interconnect density, liquid cooling system architectures, and software stack required to match rack-level performance.
Scale advantages in procurement and foundry allocations reinforce this software advantage. With Compute and Networking revenue of $193.48B in FY2026 and multi-year supply and advanced packaging commitments secured across leading foundry partners, NVIDIA secures priority access to advanced packaging nodes and high-bandwidth memory. These structural advantages sustain pricing power, as reflected in guided gross margins settling between 72% and 73% in FY2028.
NVIDIA generates substantial economic profits, evidenced by an ROIC of 87.30% compared against its moderate cost of capital of 12.6706% and WACC of 12.6133%. Consolidated operating income reached $130.39B on revenue of $215.94B in FY2026, generating an operating margin of 60.38%. This wide spread of economic returns over capital costs is protected by proprietary software architecture and rack-scale integration, preventing return erosion toward commodity hardware returns.24
Each lens answers a different valuation question. Only included lenses set Fair value; floors and diagnostics provide context.
Project SBC-adjusted EPS, discount each year's earnings at cost of equity, then add a Gordon-growth terminal value.
Under the moderate discount rate, US$152 comes from forecast-period earnings and US$167 from the long-term value, totaling US$319.
Apply a terminal P/E to the dated forward EPS path, then discount that future share value back to the analysis date.
Convert the revenue and margin path into after-tax free cash flow, discount the annual per-share amounts and terminal value at WACC, then adjust for signed net debt per current share.
Multiply the peer median PEG by this stock's forecast growth to derive an implied forward P/E, then apply it to forward EPS.
Solve for the growth rate that makes the model value equal the report-date share price.
Add depreciation to net income, subtract maintenance capex, divide by diluted shares, and apply a normalized valuation multiple.
None of the valuation methods used to calculate Fair value relies on a separate assumption about how long the competitive advantage lasts.
The highlighted cell is Fair value when every assumption sits at its base level ($374.88). Each row and column shows what Fair value becomes when one assumption moves.
Free cash flow to the firm stays positive in the first projected year at $206.8B after modeled reinvestment of $4.9B, with terminal value at 50.6% of enterprise value. A sudden drop in free cash flow is not projected unless capital spending relative to revenue, reinvestment, or gross margin resets materially.
In the Base scenario, NVIDIA delivers on management guidance with revenue growing 89.4% in FY2027 to $408.92B and 66.3% in FY2028 to $680.18B, aligning closely with sell-side consensus estimates of $682.87B. Secured multi-year packaging and foundry commitments allow fulfillment of existing backlog across Blackwell and Rubin platforms. Gross margins normalize in the 72-74% range, and operating margins settle near 58-61%. Discounting explicit cash flows at a moderate cost of equity of 12.6706% yields a fair value midpoint of $374.88.125
In the Bull scenario, continuous agentic AI inference token demand inflects faster than forecasted, driving neocloud installed capacity above 8 GW and accelerating sovereign and enterprise AI cluster investments. Revenue growth sustains at triple-digit rates in FY2027 before delivering 82.9% growth in FY2028, supported by rapid customer adoption of Vera Rubin systems. Operating margins expand toward 76-78% as high-margin proprietary software licensing attaches to NVL rack systems, driving diluted EPS compounding well above consensus targets and supporting intrinsic value reaching $586.75.5
Top-line and margin upside demonstrates accelerating Vera Rubin revenue mix and unconstrained demand exceeding supply capacity, validating the compounding trajectory toward consensus FY2028 revenue of $682.87B.
In-line execution confirms operational progress on track with management full-year expectations and guided ~70% growth for FY2028.
Revenue underperformance would indicate unexpected packaging bottlenecks or early capex digestion by top hyperscalers, requiring an adjustment of near-term growth expectations.
NVIDIA demonstrates high earnings quality characterized by strong cash conversion and clean operating margins. FY2026 operating cash flow reached $102.72B relative to net income of $120.07B and operating income of $130.39B. GAAP diluted EPS of $4.78 for FY2026 is underpinned by real cash collections rather than aggressive revenue recognition. While capex expanded to $6.04B in FY2026 to build out rack-scale system testing facilities, free cash flow generation remains abundant, supporting high-quality earnings compounding.4
Validates whether near-term shipment cadence fulfills strong market demand against supply constraints.5
Signals whether complex NVL72 rack assembly yield friction compresses profitability below normalized levels.
Determines whether multi-year foundry commitments support delivering the guided ~70% expansion.5
Tracks customer diversification outside primary hyperscalers into dedicated AI cloud infrastructure.5
Operating cash flow expanded to $102.72B in FY2026, demonstrating high conversion against net income of $120.07B. Free cash flow has compounded dramatically from $8.13B in FY2022, reaching tens of billions annually despite capital expenditures scaling to $6.04B in FY2026 (2.13x D&A) from $1.07B in FY2024. Capital intensity remains exceptionally low relative to gross cash generation, enabling self-funded capacity buildouts, advanced packaging commitments, and robust shareholder cash distributions.47
NVIDIA prioritizes reinvestment in software and architectural R&D to sustain technology leadership across Blackwell and Vera Rubin platforms. Cash returns to shareholders are executed predominantly via opportunistic share repurchases, while maintaining a modest cash dividend payout. Balance sheet liquidity remains pristine, with total debt of $38.86B representing a modest debt weight of 0.72% against a total market capitalization of $5.37T, ensuring substantial strategic flexibility.
The massive spread between ROIC and cost of capital is driven by operating margins of 60.38% on FY2026 revenue of $215.94B. High asset turns enabled by fabless production with outsourced foundry partners allow NVIDIA to generate exceptional economic value on a modest invested capital base.
ROIC of 87.30% vastly surpasses the moderate WACC of 12.6133% and cost of equity of 12.6706%.
FY2026 Compute and Networking segment delivered $193.48B in revenue and $130.14B in operating income, representing 89.6% of consolidated sales and 99.8% of operating income.
Consolidated growth and valuation are overwhelmingly driven by datacenter accelerator adoption, relegating gaming and graphics ($22.46B revenue) to secondary operational drivers.
Customer One accounted for 22.0% ($47.51B) and Customer Two accounted for 14.0% ($30.23B) of FY2026 revenue, concentrating 36.0% of revenue in two accounts.
Any capital expenditure digestion or ASIC substitution by either of these two customers poses an immediate risk to consensus forward growth trajectories.
United States revenue totaled $149.62B (69.3%) in FY2026, while China (including Hong Kong) contributed $19.68B (9.1%) and Taiwan contributed $42.35B (19.6%).
The dominant concentration of demand within domestic hyperscalers insulates aggregate financial performance against incremental export restrictions on advanced chips into China.
| 매출 | US$96,221,000,000 |
|---|---|
| 매출원가 | US$24,079,000,000 |
| 매출총이익 | US$72,142,000,000 |
| 영업비용 | US$8,408,000,000 |
| 영업이익 | US$63,734,000,000 |
| 영업외 항목 및 세금 | -US$4,046,000,000 |
| 순이익 | US$59,688,000,000 |
| 총자산 | US$206,803,000,000 |
|---|---|
| 현금 | US$10,605,000,000 |
| 유동자산 | US$125,605,000,000 |
| 총부채 | US$49,510,000,000 |
| 유동부채 | US$32,163,000,000 |
| 주주자본 | US$157,293,000,000 |
| 순이익 | US$120,067,000,000 |
|---|---|
| 영업 현금 유입 | US$129,296,000,000 |
| 감가상각비 | US$2,843,000,000 |
| 주식 기반 보상 | US$6,386,000,000 |
| 영업 현금흐름 | US$102,718,000,000 |
| 영업 현금 유출 | US$26,578,000,000 |
| 사용 가능 현금 | US$111,307,000,000 |
| 기초 현금 | US$8,589,000,000 |
| 투자 현금 유출 | US$52,228,000,000 |
| 자본적 지출 | US$6,042,000,000 |
| 기타 투자 활동 | US$46,186,000,000 |
| 재무 현금 유출 | US$48,474,000,000 |
| 자사주 매입 | US$40,086,000,000 |
| 지급 배당금 | US$974,000,000 |
| 기타 재무 활동 | US$7,414,000,000 |
| 기말 현금 | US$10,605,000,000 |
| 투자 현금흐름 | -US$52,228,000,000 |
| 재무 현금흐름 | -US$48,474,000,000 |
출처: SEC EDGAR · 제출일 2026-02-25 · SEC EDGAR
| 영업 부문별 매출 | 매출 | 비중 | 전년 대비 |
|---|---|---|---|
| Compute And Networking | US$1934.8억 | 89.6% | 66.5% |
| Graphics | US$224.6억 | 10.4% | 57% |
출처: SEC EDGAR · 제출일 2026-02-25 · SEC EDGAR
| 지역별 매출 | 매출 | 비중 | 전년 대비 |
|---|---|---|---|
| United States | US$1496.2억 | 69.3% | 93.1% |
| Taiwan | US$423.5억 | 19.6% | 79.4% |
| China Including Hong Kong | US$196.8억 | 9.1% | -21.4% |
| Other | US$43억 | 2% | -1.6% |
2026-09-19 분석에 사용된 시점 재무 데이터입니다.
| 기간 | 매출 | EBITDA | 순이익 | 영업 현금흐름 | 잉여 현금흐름 | 자본적 지출 | 감가상각비 | 주식 기반 보상 | 자사주 매입 | 지급 배당금 | 순부채 발행 | 투자 현금흐름 | 재무 현금흐름 | 기초 현금 | 기말 현금 | 현금 및 투자자산 | 총부채 | 영업이익률 | ROIC | WACC | ROIC−WACC 스프레드 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2017-01-31 | US$69.1억 | US$21.5억 | US$16.7억 | US$16.7억 | US$15억 | US$1.8억 | US$1.9억 | US$2.5억 | — | US$2.6억 | — | -US$7.9억 | US$2.9억 | — | US$17.7억 | US$68억 | US$28.2억 | 28.0% | — | 8.2% | — |
| 2018-01-31 | US$97.1억 | US$34.6억 | US$30.5억 | US$35억 | US$29.1억 | US$5.9억 | US$2억 | US$3.9억 | — | US$3.4억 | — | US$12.8억 | -US$25.4억 | US$17.7억 | US$40억 | US$71.1억 | US$20억 | 33.0% | — | 8.0% | — |
| 2019-01-31 | US$117.2억 | US$42.2억 | US$41.4억 | US$37.4억 | US$31.4억 | US$6억 | US$2.6억 | US$5.6억 | — | US$3.7억 | — | -US$41억 | -US$28.7억 | US$40억 | US$7.8억 | US$74.2억 | US$20.8억 | 32.5% | — | 9.3% | — |
| 2020-01-31 | US$109.2억 | US$34억 | US$28억 | US$47.6억 | US$42.7억 | US$4.9억 | US$3.8억 | US$8.4억 | — | US$3.9억 | — | US$61.5억 | -US$7.9억 | US$7.8억 | US$109억 | US$109억 | US$25.5억 | 26.1% | — | 8.8% | — |
| 2021-01-31 | US$166.8억 | US$56.9억 | US$43.3억 | US$58.2억 | US$46.9억 | US$11.3억 | US$11억 | US$14억 | — | US$4억 | — | -US$196.8억 | US$38억 | US$109억 | US$8.5억 | US$115.6억 | US$76억 | 27.2% | — | 6.6% | — |
| 2022-01-31 | US$269.1억 | US$113.5억 | US$97.5억 | US$91.1억 | US$81.3억 | US$9.8억 | US$11.7억 | US$20억 | — | US$4억 | — | -US$98.3억 | US$18.7억 | US$8.5억 | US$19.9억 | US$212.1억 | US$116.9억 | 37.3% | — | 7.1% | — |
| 2023-01-31 | US$269.7억 | US$59.9억 | US$43.7억 | US$56.4억 | US$38.1억 | US$18.3억 | US$15.4억 | US$27.1억 | US$100.4억 | US$4억 | US$0 | US$73.8억 | -US$116.2억 | US$19.9억 | US$33.9억 | US$133억 | US$118.6억 | 15.7% | — | 10.0% | — |
| 2024-01-31 | US$609.2억 | US$355.8억 | US$297.6억 | US$280.9억 | US$270.2억 | US$10.7억 | US$15.1억 | US$35.5억 | US$95.3억 | US$4억 | -US$12.5억 | -US$105.7억 | -US$136.3억 | US$33.9억 | US$72.8억 | US$259.8억 | US$110.6억 | 54.1% | — | 10.3% | — |
| 2025-01-31 | US$1305억 | US$861.4억 | US$728.8억 | US$640.9억 | US$608.5억 | US$32.4억 | US$18.6억 | US$47.4억 | US$337.1억 | US$8.3억 | -US$12.5억 | -US$204.2억 | -US$423.6억 | US$72.8억 | US$85.9억 | US$432.1억 | US$102.7억 | 62.4% | — | 10.9% | — |
| 2026-01-31 | US$2159.4억 | US$1445.5억 | US$1200.7억 | US$1027.2억 | US$966.8억 | US$60.4억 | US$28.4억 | US$63.9억 | US$400.9억 | US$9.7억 | US$0 | -US$522.3억 | -US$484.7억 | US$85.9억 | US$106.1억 | US$625.6억 | US$114.1억 | 60.4% | — | 12.7% | — |
NVIDIA Corporation (NVDA) trades at 24.2x forward earnings against a peer median of 20.7x. Its EV/EBITDA of 26.5x compares with a peer median of 30.3x. The comparison uses 4 peers: Broadcom Inc., Micron Technology Inc., Advanced Micro Devices Inc. and Analog Devices Inc., as of 2026-09-18 to 2026-09-19.
Semiconductor industry capex-to-revenue stands at 4.4% with capex-to-depreciation at 1.27x. NVIDIA's capital expenditures grew from $1.07B in FY2024 to $6.04B in FY2026 (2.13x D&A), supporting system testing and advanced packaging infrastructure for rack-scale systems.4
In an extended bull cycle, the inflection from prompt-based AI into continuous autonomous agentic inference drives an exponential surge in token consumption, compelling hyperscalers, neoclouds, and sovereigns to compound datacenter capex above 50% annually through FY2028. NVIDIA's secured multi-year foundry and packaging commitments allow it to capture the majority of industry spending, sustaining revenue growth near guided ~70% levels with operating margins expanding toward 62%.
In a downcycle or digestion phase, hyperscalers pause physical datacenter buildouts following initial foundation cluster completions due to power interconnect bottlenecks and slower enterprise ROI monetization. Slower spending by the top two customers (36.0% FY2026 revenue share) decelerates industry accelerator shipments, leading to component inventory build and pricing concessions that compress gross margins toward 70%.
NVIDIA occupies the apex positioning within accelerated computing infrastructure, commanding dominant market share across AI accelerators and high-speed datacenter networking fabrics. By pairing compute GPUs with proprietary NVLink switches, Mellanox networking, and full-stack software runtimes, NVIDIA captures full rack-level system value rather than commoditized chip margins. Furthermore, balanced geographic exposure—with U.S. revenue representing 69.3% ($149.62B) versus China at 9.1% ($19.68B) —positions NVIDIA to absorb international regulatory restrictions without structural disruption.56
High conviction is underwritten by NVIDIA's wide economic moat, exceptional ROIC of 87.30% exceeding capital costs of 12.6706%, and multi-year visibility supported by secured foundry commitments. A 5-7% target weight balances dominant technology leadership against customer concentration risk (36% top-two customers).256
Portfolio risk is managed by enforcing a maximum allocation cap of 7.0%, limiting exposure to broader semiconductor sector volatility (raw beta of 2.2478). Cross-correlation with megacap cloud hyperscalers is monitored quarterly, and rebalancing is executed automatically upon reaching defined take-profit and checkpoint thresholds.
At $222.27, NVIDIA trades at a 40.7% discount to house Fair value of $374.88, anchored by consensus FY2028 EPS of $15.68 and structural free cash flow compounding.14
Management guided ~70% revenue growth for FY2028 supported by post-earnings upward revisions to $682.87B consensus revenue.15
Customer concentration across top two hyperscalers (36.0%) creates vulnerability to capex digestion, though current 24.15x forward P/E already provides downside protection relative to peer medians.46
NVIDIA is rated Strong Buy with a house Fair value of $374.88 per share, offering 68.7% upside from the analysis date price of $222.27. The recommendation is underwritten by structural moat advantages in CUDA, multi-year capacity reservations, and consensus FY2028 EPS resetting to $15.68.145
While customer concentration among cloud hyperscalers (36% combined share) introduces capex digestion risk, non-hyperscaler enterprise, sovereign AI, and neocloud expansion mitigate downside severity and support compounding free cash flow.
Capture your own thesis, position size, and follow-up questions. Stays on this device, only you can see it.
The records this analysis drew on, as they stood when it was run.
Not every cited input appears here: internal calculations and intermediate model outputs have no external record, and some citations could not be matched to one.
