The artificial intelligence revolution has created the most valuable company in history, and NVIDIA sits at its epicenter. With a $5.5 trillion market capitalization and data center revenue that has tripled in two years, the question facing investors is not whether NVIDIA dominates AI infrastructure—that is established fact—but whether the stock price has already captured the opportunity ahead. This analysis argues that despite the extraordinary run, NVIDIA’s Blackwell architecture cycle, combined with $700 billion in hyperscaler capital expenditure commitments and a forward P/E below 25x, creates a compelling case for continued outperformance.
Three Key Investment Points:
1. Blackwell Demand Exceeds All Prior Product Cycles: With 3.6 million units backordered and systems sold out through mid-2026, NVIDIA’s next-generation architecture is experiencing unprecedented demand. CEO Jensen Huang describes appetite as “off the charts,” backed by $1 trillion in confirmed orders through 2027.
2. Data Center Moat Continues to Widen: NVIDIA commands 80-85% of the AI accelerator market by revenue, and despite AMD and custom silicon competition, absolute revenue continues growing as the total addressable market expands faster than share erosion.
3. Valuation Disconnect at Forward Multiples: At 24.5x forward earnings with 65%+ revenue growth, NVIDIA trades at a significant discount to its growth rate. Bank of America’s $320 target implies 42% upside, while consensus at $275 suggests 22% from current levels.
This article provides a comprehensive analysis of NVIDIA’s business model, competitive positioning, financial trajectory, and valuation framework to help investors make informed decisions about the world’s most important technology company.
—
1. Company Overview
NVIDIA Corporation, founded in 1993 by Jensen Huang, Chris Malachowsky, and Curtis Priem, has transformed from a graphics card manufacturer into the dominant force in artificial intelligence computing infrastructure. Headquartered in Santa Clara, California, the company designs and sells graphics processing units (GPUs), system-on-chip units, and related software that power everything from gaming PCs to the world’s most advanced AI data centers.
Business Model: How NVIDIA Generates Revenue
NVIDIA operates a fabless semiconductor model, designing chips while outsourcing manufacturing to foundry partners—primarily Taiwan Semiconductor Manufacturing Company (TSMC). This capital-light approach allows NVIDIA to focus on architecture innovation and software ecosystem development while maintaining industry-leading gross margins above 70%.
The company generates revenue through three primary channels:
1. Direct Sales to Hyperscalers: Microsoft, Amazon, Google, and Meta collectively represent the largest customer base, purchasing GPU systems directly for their cloud infrastructure and internal AI workloads.
2. OEM and System Integrator Partnerships: Dell, HPE, Lenovo, and other hardware vendors integrate NVIDIA GPUs into enterprise server solutions.
3. Software and Licensing: CUDA (Compute Unified Device Architecture) remains free, but enterprise AI software suites including NVIDIA AI Enterprise command subscription revenue.
Revenue Breakdown by Segment (FY2026)
Segment Revenue % of Total YoY Growth Data Center $193.7B 89.7% +68% Gaming $24.1B 11.2% +15% Professional Visualization $2.1B 1.0% +12% Automotive $1.9B 0.9% +55% OEM & Other $0.6B 0.3% -8% Total $215.9B 100% +65.5%
Key Customers and Market Position
NVIDIA’s top customers include the “Magnificent Seven” hyperscalers—Microsoft, Amazon, Google, Meta, Apple, and Tesla—plus sovereign AI initiatives from nations including Saudi Arabia, UAE, and Japan. The company estimates that its top four cloud customers alone represent over 40% of data center revenue.
In the AI accelerator market, NVIDIA holds an estimated 80-85% share by revenue, down from 92% in 2023 as AMD and custom silicon alternatives gain traction. However, this share decline masks absolute growth: NVIDIA’s data center revenue grew from $47 billion in FY2025 to $194 billion in FY2026—a 313% increase in two years.
Ownership Structure
Institutional investors hold approximately 65% of NVIDIA shares, with Vanguard (8.7%), BlackRock (7.2%), and State Street (4.1%) as the largest holders. Founder and CEO Jensen Huang maintains a 3.5% stake worth approximately $190 billion, making him one of the world’s wealthiest individuals. Huang’s ownership alignment provides strong governance incentive, and his technical background—he holds a master’s degree in electrical engineering from Stanford—gives him deep credibility with customers and engineers alike.
—
2. Industry Analysis
2-1. Market Size & Growth Trajectory
The AI infrastructure market represents one of the largest technology investment cycles in history. According to Bank of America’s May 2026 research, the total addressable market (TAM) for AI data centers will reach $1.7 trillion by 2030, up from approximately $400 billion in 2025. This implies a compound annual growth rate (CAGR) of 33% over five years—exceptional for a market of this scale.
Several factors drive this extraordinary expansion:
Cloud Infrastructure Spending Acceleration: The four largest cloud providers—Microsoft Azure, Amazon Web Services, Google Cloud, and Meta—are projected to spend a combined $700 billion on capital expenditures in 2026, with the vast majority directed toward AI infrastructure. Microsoft alone has committed $80 billion in AI data center investments for fiscal 2025, while Meta’s AI infrastructure spending has grown from $28 billion in 2024 to over $60 billion in 2026.
Enterprise AI Adoption: Beyond hyperscalers, enterprise customers are rapidly deploying AI infrastructure. Goldman Sachs estimates that enterprise AI spending will grow from $50 billion in 2024 to $250 billion by 2028 as companies implement generative AI for customer service, code generation, content creation, and workflow automation.
Sovereign AI Initiatives: Nations are treating AI capability as strategic infrastructure. Saudi Arabia’s Project Transcendence committed $100 billion to AI infrastructure through 2030, while the UAE’s G42 and Japan’s national AI strategy represent multi-billion-dollar GPU procurement programs. NVIDIA estimates sovereign AI represents 15% of data center revenue and is growing faster than commercial demand.
The industry sits firmly in the acceleration phase of its adoption curve. Unlike previous technology cycles where demand normalized after initial deployment, AI workloads exhibit compound characteristics: larger models require exponentially more compute, inference demand scales with user adoption, and new applications continuously emerge. This creates what NVIDIA CFO Colette Kress calls “sustained demand visibility”—the company can see $500 billion in Blackwell and Rubin revenue from the start of calendar 2025 through the end of calendar 2026.
2-2. Structural Growth Drivers
Three structural forces ensure AI infrastructure demand extends beyond the current investment cycle:
Model Scaling Laws Continue to Hold: Despite periodic concerns about diminishing returns from larger models, empirical evidence demonstrates that performance continues improving with scale. GPT-5, Claude 4.6, and Gemini 2.0 all show capability gains versus predecessors, and training clusters have grown from thousands to hundreds of thousands of GPUs. The next generation of frontier models will require million-GPU clusters, representing 10x growth in training infrastructure even if model releases slow from annual to biennial cadence.
Inference Demand Growing Faster Than Training: While training garners headlines, inference workloads—running trained models to generate outputs—represent the majority of GPU hours at hyperscale. Every ChatGPT conversation, every Copilot code completion, every AI-generated image requires inference compute. As AI applications proliferate, inference demand scales with user adoption rather than model development cycles. NVIDIA estimates inference now represents over 50% of data center GPU utilization, up from 30% in 2023.
Physical AI Emergence: The next frontier extends beyond language models to robotics, autonomous vehicles, and industrial automation. NVIDIA’s partnership with Tesla, BMW, and manufacturing companies positions its GPUs as the training and inference platform for physical AI systems. Jensen Huang has stated that physical AI—training robots to interact with the real world—will represent the largest AI market within a decade, potentially exceeding language model infrastructure.
Energy and Efficiency Requirements: As AI workloads consume increasing electricity, demand shifts toward newer architectures offering superior performance per watt. Blackwell delivers 25x the performance of Hopper at equivalent power, creating an upgrade cycle even among customers with existing GPU fleets. Hyperscalers face both economic pressure (electricity costs) and regulatory requirements (carbon commitments) that favor the latest architectures.
2-3. Competitive Landscape
NVIDIA faces competition from three directions: AMD’s GPU accelerators, custom silicon from hyperscalers, and Intel’s struggling Gaudi lineup. Despite competitive pressure, NVIDIA’s position has strengthened rather than weakened.
Competitor Comparison Table:
Company 2025 AI Revenue 2026E AI Revenue Market Share Key Products Primary Advantage NVIDIA $130.5B $215.9B 80-85% H100, H200, Blackwell CUDA ecosystem, training dominance AMD $6.5B $15.0B 5-7% MI300X, MI325X Price/performance on inference Google TPU N/A N/A 6-8%* TPUv5, TPUv6 Internal workload optimization AWS Trainium N/A N/A 2-3% Trainium 2 Cost arbitrage on AWS inference Intel $1.2B $0.8B <1% Gaudi 3 None (execution failure)
*Google TPU share measured by deployed FLOPS rather than revenue
Why NVIDIA Maintains Dominance:
The CUDA software ecosystem represents NVIDIA’s deepest competitive moat. Over 4 million developers have trained on CUDA, and enterprise AI software stacks are built on NVIDIA libraries including cuDNN, TensorRT, and NCCL. Switching costs are substantial: rewriting inference pipelines for AMD ROCm or custom silicon requires months of engineering effort and introduces execution risk that enterprises avoid when possible.
Training workloads remain almost exclusively NVIDIA territory. The largest language models—GPT-5, Claude, Gemini—train on NVIDIA hardware because alternatives cannot match performance at scale. While AMD’s MI300X shows competitive inference benchmarks, training clusters require proven reliability at 100,000+ GPU scale that only NVIDIA has demonstrated.
Network effects compound NVIDIA’s advantage. The company’s NVLink and InfiniBand networking infrastructure optimizes multi-GPU communication, while competitors rely on standard Ethernet. At training scale, network bandwidth becomes the bottleneck, and NVIDIA’s integrated system approach—GPU, networking, and software—delivers 2-3x effective throughput versus disaggregated alternatives.
—
3. Economic Moat Analysis
NVIDIA possesses multiple overlapping competitive advantages that constitute one of the widest economic moats in the technology industry. Understanding these advantages—and their durability—is essential for long-term investment analysis.
Moat Type 1: Switching Costs (CUDA Ecosystem Lock-In)
NVIDIA’s CUDA platform represents the most significant source of competitive advantage. Launched in 2007, CUDA provides the programming model, libraries, and tools that developers use to write GPU-accelerated software. The ecosystem’s scale is extraordinary:
– 4+ million developers trained on CUDA
– 900+ GPU-accelerated applications across industries
– 600+ universities teaching CUDA in curriculum
– 50+ AI frameworks natively supporting CUDA
The switching costs manifest at multiple levels. Individual developers have invested years learning CUDA optimization techniques. Enterprises have built proprietary inference pipelines atop CUDA libraries. AI startups have written millions of lines of CUDA-specific code. Research institutions have standardized on NVIDIA hardware for reproducibility.
AMD’s ROCm platform has made progress, but adoption remains limited. Major AI frameworks (PyTorch, TensorFlow) support ROCm, yet performance parity requires optimization work that most organizations avoid. The practical reality: switching from NVIDIA to AMD requires 6-12 months of engineering effort for a typical enterprise AI deployment, plus ongoing maintenance burden as frameworks evolve.
Moat Type 2: Network Effects (Developer Ecosystem Compounding)
CUDA’s competitive moat exhibits network effect characteristics that strengthen over time. As more developers use CUDA, more libraries and tools emerge. As the ecosystem grows richer, more developers join. This flywheel has operated for nearly two decades, accumulating advantages that new entrants cannot replicate.
The network effects extend to enterprise customers. Companies share CUDA optimization techniques, performance benchmarks, and deployment patterns. Hiring GPU engineers means hiring CUDA engineers—the talent pool is NVIDIA-specific. IT organizations prefer NVIDIA because support ecosystems, consulting expertise, and reference architectures exist at scale.
NVIDIA actively cultivates these network effects through developer relations, GTC conferences, and academic partnerships. The company estimates it has invested over $50 billion in R&D since 2019, with substantial allocation toward software ecosystem development. This investment creates sustainable advantage: competitors would need to replicate not just hardware performance but two decades of ecosystem development.
Moat Type 3: Cost Advantage (At-Scale Manufacturing Leverage)
NVIDIA’s scale provides manufacturing cost advantages that smaller competitors cannot match. As the largest customer of TSMC’s most advanced nodes (N4, N3), NVIDIA receives favorable pricing, priority allocation during shortages, and early access to next-generation processes.
The company’s wafer commitments—reportedly exceeding $25 billion annually with TSMC alone—secure capacity that competitors struggle to obtain. During the 2020-2023 chip shortage, NVIDIA maintained supply while AMD and Intel faced constraints. This reliability matters to hyperscaler customers planning multi-year infrastructure buildouts.
Scale advantages extend to system design. NVIDIA’s networking division (from the Mellanox acquisition) provides NVLink and InfiniBand interconnects that optimize multi-GPU communication. Competitors must either source networking from third parties or develop capabilities in-house, adding cost and complexity.
Moat Durability Assessment
NVIDIA’s moat faces identifiable threats that could erode competitive position over 5-10 years:
Custom Silicon Risk: Google, Amazon, Microsoft, and Meta all develop internal AI accelerators. If custom silicon achieves training parity at scale, hyperscaler share could shift substantially. However, internal chip programs have consistently underperformed expectations—Google’s TPU remains confined to Google workloads, AWS Trainium has not achieved broad adoption, and Meta abandoned its training chip program.
AMD Execution Improvement: AMD under CEO Lisa Su has executed effectively, and MI300X shows competitive inference performance. If AMD solves its software ecosystem challenges and achieves training reliability at scale, enterprise share could erode more rapidly than projected.
Architectural Disruption: Longer-term, alternative computing paradigms—neuromorphic chips, optical computing, quantum acceleration—could challenge GPU dominance. However, no alternative approaches commercial viability within the forecast horizon.
On balance, NVIDIA’s moat appears durable for the next five years. The CUDA ecosystem switching costs, network effects, and scale advantages create compounding barriers that increase rather than erode over time. Custom silicon represents the greatest threat, but hyperscaler programs have not demonstrated capability to displace NVIDIA at training scale.
—
4. Financial Analysis
NVIDIA has delivered the most extraordinary financial performance in semiconductor history, with revenue tripling and earnings growing sixfold over two fiscal years. Understanding the sustainability of this trajectory requires examining both historical trends and forward-looking drivers.
Revenue and Profitability Trends
Metric FY2024 FY2025 FY2026 FY2027E Revenue $60.9B $130.5B $215.9B $280.0B YoY Growth +122% +114% +65% +30% Data Center $47.5B $115.2B $193.7B $255.0B Gross Margin 72.7% 74.6% 73.0% 74.0% Operating Income $32.9B $81.2B $138.0B $180.0B Operating Margin 54.1% 62.2% 63.9% 64.3% Net Income $29.8B $72.9B $122.8B $160.0B EPS (Diluted) $1.19 $2.94 $4.92 $6.40
The FY2026 results demonstrate continued hypergrowth despite an already massive base. Data center revenue of $193.7 billion grew 68% year-over-year, driven by Hopper-to-Blackwell transition and expanding hyperscaler deployments. Gross margins compressed slightly to 73% due to Blackwell ramp costs and China export restrictions, but management guides recovery to mid-70% levels as production scales.
Q1 FY2027 Performance
The most recent quarter (Q1 FY2027) showed continued strength:
– Revenue: $44.06 billion (+69.2% YoY)
– Data Center: $39.11 billion (89% of total)
– Gaming: $3.76 billion (+42% YoY, record)
– EPS: $0.81 (beat $0.75 consensus by 8%)
– Gross Margin: 70.6% (impacted by $4.5B H20 charge)
The H20 charge relates to U.S. government export restrictions on China-bound chips, requiring NVIDIA to write down inventory and purchase commitments. Management absorbed this one-time impact while guiding Q2 revenue to $45 billion—implying the underlying business continues accelerating.
Key Operating Metrics
Beyond standard financials, several metrics illuminate business quality:
Backlog and Visibility: NVIDIA reports $500+ billion in confirmed customer orders extending through end of calendar 2026. Blackwell systems are sold out through mid-year 2026, with 3.6 million units backordered across hyperscalers and sovereign customers.
Data Center ASP Growth: Average selling price per data center GPU continues rising as customers purchase higher-end configurations. Blackwell systems command premiums versus Hopper, driving revenue growth even at flat unit volumes.
Software Attach Rate: NVIDIA AI Enterprise software revenue has grown from negligible to an estimated $2+ billion annually, with enterprise customers increasingly purchasing software subscriptions alongside hardware.
Balance Sheet Strength
Metric FY2026 Cash & Short-Term Investments $43.2B Total Debt $8.5B Net Cash $34.7B Free Cash Flow $95.3B FCF Margin 44.1%
NVIDIA generates cash at extraordinary rates—nearly $100 billion in free cash flow during FY2026. The company returned $27 billion to shareholders through buybacks and dividends while maintaining a fortress balance sheet. This financial strength enables continued R&D investment ($12 billion annually), strategic acquisitions, and shareholder returns regardless of economic conditions.
—
5. Valuation
NVIDIA’s valuation presents the classic growth-stock paradox: headline multiples appear elevated, but growth rates justify premium pricing. This section provides a framework for assessing fair value across multiple scenarios.
Current Valuation Metrics
Metric Current 5-Year Average Sector Median P/E (TTM) 45.7x 52.3x 25.1x P/E (Forward) 24.5x 35.8x 18.7x EV/EBITDA 35.2x 42.1x 15.3x EV/Revenue 24.8x 18.5x 4.2x PEG Ratio 0.71x 1.05x 1.42x
The forward P/E of 24.5x stands out: at current prices, NVIDIA trades at a 35% discount to its five-year average multiple despite continued hypergrowth. The PEG ratio of 0.71x—below 1.0—suggests the stock is undervalued relative to its growth rate.
DCF Valuation Framework
Using discounted cash flow analysis with the following assumptions:
Base Case Assumptions:
– Revenue CAGR: 25% (FY2027-FY2032)
– Terminal growth: 3%
– Operating margin: 62% (stable)
– WACC: 10%
– Tax rate: 15%
DCF Output:
– Intrinsic value per share: $285
– Current price: $225.32
– Implied upside: 26.5%
Bull Case Assumptions (adjusted):
– Revenue CAGR: 30% (continued hyperscaler acceleration)
– Operating margin: 65% (software mix improvement)
– Lower WACC: 9% (reduced risk premium as AI matures)
Bull Case Output:
– Intrinsic value: $380
– Implied upside: 69%
Bear Case Assumptions:
– Revenue CAGR: 15% (custom silicon accelerates, China revenue loss persists)
– Operating margin: 55% (competitive pressure)
– Higher WACC: 12% (multiple compression)
Bear Case Output:
– Intrinsic value: $180
– Implied downside: 20%
Analyst Consensus Comparison
Firm Rating Price Target Upside Bank of America Buy $320 +42% Wells Fargo Overweight $315 +40% Morgan Stanley Overweight $280 +24% Consensus (37 analysts) Strong Buy $275.62 +22%
Bank of America’s $320 target, raised on May 13, 2026, reflects their $1.7 trillion 2030 AI data center TAM estimate and confidence in NVIDIA’s ability to capture the majority of this spending. The firm argues that Blackwell production scaling in H2 2026 will drive reacceleration after Q1’s export control impact.
Investment Thesis: Why $275+ Is Achievable
The path to $275 (consensus target) requires:
– FY2028 EPS of $7.50 (vs. FY2027E of $6.40)
– Forward P/E compression to 37x (below current 45.7x TTM)
– Continued execution on Blackwell ramp
This scenario assumes no multiple expansion and EPS growth decelerating to 17%—conservative relative to recent trends. The bull case to $320+ requires sustained 25%+ EPS growth and stable multiples, supported by physical AI emergence and inference scaling.
Scenario Analysis Summary
Scenario Probability Target Price Return Bull Case 25% $380 +69% Base Case 55% $285 +26% Bear Case 20% $180 -20% Probability-Weighted $277 +23%
The probability-weighted expected return of 23% supports a constructive investment stance, with asymmetric upside if AI infrastructure spending exceeds consensus expectations.
—
6. Risk Factors
Risk 1: China Export Controls and Geopolitical Escalation
U.S. government export restrictions present the most significant near-term risk to NVIDIA’s business. The April 2026 licensing requirement for H20 chips into China already cost $4.5 billion in Q1 charges, with management absorbing an estimated $8 billion annual revenue impact.
The risk extends beyond current restrictions. Further escalation—potential bans on all AI chips to China, sanctions on technology transfer, or retaliatory Chinese actions—could reduce revenue by 15-20% in a severe scenario. China represented approximately 17% of NVIDIA’s revenue in FY2025 before restrictions intensified.
Mitigation: NVIDIA has diversified manufacturing beyond Taiwan (Arizona fab partnership with TSMC) and expanded non-China sovereign AI programs to offset lost China demand. Management guided that Q1 absorbed the majority of H20 impact, suggesting limited incremental exposure.
Risk 2: Custom Silicon Displacement at Scale
Hyperscaler custom chip programs represent the greatest competitive threat to NVIDIA’s long-term dominance. Google’s TPU, Amazon’s Trainium, Microsoft’s Maia, and Meta’s MTIA all target workload-specific acceleration that could reduce GPU requirements.
If custom silicon achieves training parity, hyperscalers could shift substantial capital expenditure internally. A scenario where the top four cloud customers reduce NVIDIA purchases by 30% would impact approximately $60 billion in annual revenue.
Mitigation: Custom silicon programs have consistently underdelivered versus expectations. Google’s TPU remains confined to Google workloads, Trainium adoption is limited, and Meta abandoned its training chip. NVIDIA’s integrated platform approach—combining GPU, networking, and software—provides system-level advantages that single-purpose chips cannot match.
Risk 3: AI Investment Cycle Deceleration
The current AI infrastructure buildout could decelerate if: (a) model scaling laws exhibit diminishing returns, (b) enterprise AI adoption disappoints, (c) economic downturn constrains capital expenditure, or (d) regulatory action limits AI deployment.
Historical precedent shows technology investment cycles can reverse sharply. The 2000 dot-com collapse and 2022 crypto winter demonstrate that even transformative technologies experience boom-bust dynamics.
Mitigation: AI infrastructure exhibits structural differences from prior bubbles. The technology delivers measurable productivity gains (Microsoft reports 55% faster coding with Copilot), and hyperscaler spending reflects revenue-generating services rather than speculative positioning. Additionally, NVIDIA’s diversification into automotive, gaming, and industrial markets provides ballast if data center growth moderates.
—
7. Conclusion & Exit Plan
Investment Rating: Strong Buy
NVIDIA remains the highest-conviction holding in the AI infrastructure theme. The company’s dominant market position, expanding product portfolio, and execution track record justify continued overweight positioning despite the extraordinary price appreciation.
Entry Price Range
– Aggressive Entry: Current price ($225.32)
– Conservative Entry: Pullback to $200 (-11%)
– Opportunistic Entry: Pullback to $180 (-20%, bear case support)
The current price offers attractive risk/reward given the 22-42% upside to analyst targets. Investors seeking lower entry points may wait for volatility around earnings releases or geopolitical headlines, though waiting risks missing continued appreciation.
Exit Conditions
Target Achieved: Begin trimming at $275 (base case target), sell 50% of position at $320 (bull case target). Above $350, reduce to 2% portfolio weight regardless of fundamentals.
Fundamental Break: Exit position if any of the following occur:
– Data center revenue growth declines below 20% YoY for two consecutive quarters
– Gross margin falls below 65% without clear recovery path
– Market share erosion accelerates below 70% (vs. current 80-85%)
– Custom silicon achieves training parity at 100,000+ GPU scale
Time-Based: Reassess thesis in Q1 2027 (after full-year FY2027 results) and Q1 2028 (Rubin architecture launch).
Summary Table
Item Detail Company NVIDIA Corporation (NVDA) Current Price $225.32 Target Price (Base) $275.25 Target Price (Bull) $400.00 Target Price (Bear) $205.00 Upside (to Base) +22% Rating Strong Buy Key Thesis Blackwell demand, AI infrastructure dominance, forward P/E discount to growth Main Risk China export controls, custom silicon competition
—
8. What Changed Since Last Analysis
This represents NVIDIA’s inaugural comprehensive analysis on mybestinvesting.co.kr. The holding thesis has been tracked in Think Tank since 2024, when an earlier Q1 2024 earnings review was published. This section documents how the investment case has evolved over the intervening two years.
Original Thesis Elements (2024)
At the time of our Q1 2024 coverage, the investment thesis centered on:
1. AI Data Center Leadership: NVIDIA’s H100 GPU dominated training workloads, with hyperscalers deploying at unprecedented scale. This thesis has strengthened dramatically—data center revenue grew from $18 billion quarterly to $39 billion, and market share has remained above 80%.
2. Long-term Multibagger Potential: The original thesis characterized NVIDIA as a “return-of-principal-then-hold-forever” position. With +1,816% total return since initial purchase, this characterization has been validated. The position has moved from aggressive accumulation to profit-taking and position management.
3. Monitoring Data Center Growth: The key watch item was whether data center momentum would sustain. FY2025 and FY2026 results exceeded even optimistic scenarios, with the segment growing 68% and 114% in successive years.
New Investment Ideas Emerged
Since the original analysis, several new thesis elements have emerged:
Blackwell Architecture Cycle (New): The transition from Hopper to Blackwell creates another upgrade wave, with 3.6 million units backordered and systems sold out through mid-2026. This was not contemplated in the 2024 analysis and represents incremental upside.
Sovereign AI Demand (New): Government-led AI infrastructure programs in Saudi Arabia, UAE, Japan, and Europe have emerged as a material revenue driver, now representing 15% of data center sales. This diversification reduces hyperscaler concentration risk.
Inference Scaling (Evolved): The 2024 thesis focused on training infrastructure. Inference has since become 50%+ of GPU utilization, creating a more sustainable demand profile that scales with AI application adoption rather than model development cycles.
Risks Not Present in Prior Analysis
H20 Export Restrictions: The April 2026 licensing requirement was not contemplated in 2024, when China represented 22% of revenue. The $8 billion annual impact, while absorbed, represents a structural change in the geographic revenue mix.
Custom Silicon Acceleration: Google, Amazon, Microsoft, and Meta have all accelerated internal chip programs since 2024. While none have achieved training parity, competitive intensity has increased.
—
9. Current Assessment
Position Performance Summary
Metric At Original Analysis Current Change Entry Price (avg) ~$12.00* N/A — Analysis Date Price $88.50 (May 2024) $225.32 +155% Current Price — $225.32 — Total Return +637% (May 2024) +1,816% +1,179pp Position Weight 4.5% 2.3% -2.2pp
*Estimated based on +1,816% total return
The position weight has decreased from 4.5% to 2.3% through profit-taking rather than underperformance. This reflects disciplined position management—taking gains as the stock appreciated while maintaining meaningful exposure.
Target Achievement Assessment
Target Value Current Price Status Base Case $275.25 $225.32 82% achieved Bull Case $400.00 $225.32 56% achieved Bear Case $205.00 $225.32 Above support
The current price sits between bear case support ($205) and base case target ($275), suggesting neither downside protection concerns nor immediate profit-taking triggers. The thesis remains intact.
Holding Stance
Active — Maintain Position: The thesis is intact, Blackwell demand exceeds expectations, and forward valuation offers reasonable upside. No impairment conditions have been triggered, and review deadline is May 2027.
The 2.3% position weight is appropriate for a stock that has appreciated 18x—large enough to benefit from continued upside, small enough to avoid concentration risk.
—
10. Revised Price Target & Valuation
Updated Assumptions vs. Original
Assumption Previous (May 2024) Current Driver FY2025E Revenue $80B $130.5B (actual) Beat by 63% FY2026E Revenue $110B $215.9B (actual) Beat by 96% FY2027E Revenue — $280B New estimate Data Center Share 85% 80-85% AMD/custom gains Terminal Margin 60% 64% Better than expected
Revised Price Targets
Scenario Previous Target Revised Target Change Key Driver Base Case $275.25 $285.00 +4% Higher FY2028 EPS on Blackwell Bull Case $400.00 $400.00 — Unchanged, achievable if AI accelerates Bear Case $205.00* $200.00 -2% Export control risk increased
*Previous targets from holding-theses.md consensus-based estimate (May 2026)
The modest base case revision reflects confidence in FY2028 earnings trajectory. Blackwell production scaling should drive 25-30% EPS growth, supporting a $285 fair value at 38x forward earnings.
The bull case remains at $400, representing ~42x FY2028 EPS of $9.50 in an acceleration scenario. This requires physical AI emergence, inference demand exceeding forecasts, and sustained hyperscaler spending.
The bear case tightens slightly to $200 from $205, reflecting incremental export control risk. At $200, NVIDIA would trade at 25x FY2028 EPS—a floor valuation implying growth normalization.
Analyst Consensus Comparison
Current analyst consensus ($275.62) aligns with our base case. Bank of America’s $320 target represents our bull-case pathway. We agree with BofA’s $1.7 trillion TAM projection but apply a higher discount rate given execution risk.
—
11. Updated Exit Plan
Recommended Stance
Continue Holding — Maintain 2.3% Weight
The thesis is intact, valuation offers upside, and no impairment triggers have been met. Position weight is appropriate for a multi-bagger that has moved through the return-of-principal phase.
Exit Strategy by Price Level
Price Action Rationale $275+ Trim 25% Base case target achieved $320+ Trim additional 25% Bull case partial achievement $400+ Reduce to 1% weight Full bull case, maintain token exposure <$180 Re-evaluate Bear case breach, assess fundamentals
Updated Impairment Conditions
Exit position (or reduce to token) if:
1. Data Center Growth Collapse: Revenue growth <15% YoY for two consecutive quarters without clear recovery catalyst
2. Margin Erosion: Gross margin <65% for two quarters, indicating competitive pressure or mix deterioration
3. Market Share Acceleration: Share drops below 70% with AMD/custom gaining traction in training workloads
4. China Escalation: Complete ban on all China AI chip sales, representing >15% revenue loss
5. Management Change: Jensen Huang departure without credible succession plan
Next Review Date
November 2026 — After FY2027 Q2 earnings (Blackwell full-quarter contribution) and mid-year update on export control landscape.
Summary Recommendation
For current holders: Maintain position at current weight. The risk/reward at $225 with $275-$320 upside targets and $180-$200 downside support remains attractive. Trim at base case achievement, accelerate trimming above bull case. No action required at current levels beyond routine monitoring.
—
Disclaimer
This article is for informational purposes only and does not constitute investment advice. All data sourced from public filings, analyst reports, company announcements, and financial news as of May 17, 2026. The author and mybestinvesting.co.kr may hold positions in securities discussed. Past performance does not guarantee future results. Invest at your own discretion and consult a qualified financial advisor before making investment decisions.
Sources:
– [NVIDIA Q1 2026 Earnings](https://247wallst.com/companies/nvda/earnings/2026/Q1)
– [Bank of America NVIDIA Price Target](https://247wallst.com/investing/2026/05/13/bofa-hikes-nvidia-price-target-to-320-on-massive-1-7-trillion-ai-data-center-forecast/)
– [NVIDIA Blackwell Architecture](https://www.nvidia.com/en-us/data-center/technologies/blackwell-architecture/)
– [NVIDIA Q4 FY2026 Results](https://fortune.com/2026/02/25/nvidia-nvda-earnings-q4-results-jensen-huang/)
– [NVIDIA Revenue History](https://www.macrotrends.net/stocks/charts/NVDA/nvidia/revenue)
– [AI Chip Market Share Analysis](https://siliconanalysts.com/analysis/nvidia-ai-accelerator-market-share-2024-2026)
함께 읽으면 좋은 글
- Microsoft Azure AI Investment Thesis: 40% Cloud Growth and $37 Billion AI Revenue Signal a Generational Platform Shift
- Apple’s iPhone 17 AI Supercycle and $31 Billion Services Record: Why the World’s Most Valuable Company Still Has 35% Upside
- [May 2026 Reanalysis] Rocket Lab Exceeds Bull Case Target After Record Q1: What Investors Should Do Now
- [2026 May Reanalysis] AMD Q1 2026 Earnings Crush Expectations: Data Center Revenue Surges 57% as MI300 Momentum Accelerates Toward $530 Price Target
- Shopify Q1 2026 AI Commerce Revolution: Why the $100 Billion GMV Milestone and 13x AI Order Growth Create a Generational Entry Point