AI Beneficiary Stress Test: Customer Concentration vs Margin Momentum (AVGO, NVDA, ANET, VRT)
AVGO's top 5 customers account for 40% of revenue. NVDA's largest customer is 22%. ANET depends on Microsoft (20%) and Meta (15%). Standard AI stock screens show none of this. Here's our 2-signal framework for stress-testing AI beneficiaries.
AI Beneficiary Stress Test: Customer Concentration vs Margin Momentum
Last Updated: December 22, 2025 Data Currency: Q3 2025 10-Q filings. AVGO, NVDA, ANET, VRT
Standard screens miss the concentration risk hiding in AI stock SEC filings. We analyzed 4 AI infrastructure beneficiaries using a 2-signal framework: customer concentration (risk dimension) and 8-quarter margin momentum (quality dimension).
Key Findings:
- AVGO: 40% concentration + +6.33pp margin improvement = HIGH RISK, HIGH REWARD
- NVDA: 22% single customer + margin pressure = ELEVATED RISK, EXCEPTIONAL ROIC
- ANET: 35% hyperscaler dependence + flat margins = HIGH RISK, EXCEPTIONAL ROIC
- VRT: No concentration + steady improvement = LOW RISK, QUALITY PROFILE
- Broadcom (AVGO): Top 5 customers account for 40% of revenue; gross margin improved +6.33pp over 8 quarters
- NVIDIA (NVDA): Customer A represents 22% of total revenue; 4 customers exceed 10% threshold
- Arista Networks (ANET): Microsoft 20%, Meta Platforms 15% of revenue; ROIC accelerated +54.54pp
- Vertiv (VRT): No customer concentration disclosed; "diversification strategy to avoid over concentration"
- AVGO ROIC: Improved +54.91pp over 8 quarters (72.7% current)
- ANET ROIC: 160.8% current, highest capital efficiency among peers
Sources: Company 10-Q/10-K filings, SEC EDGAR
MetricDuck automatically tracks customer concentration and margin trends across 1,000+ companies. Compare All 4 AI Beneficiaries →
Why Do Standard AI Stock Screens Miss Concentration Risk?
Everyone's screening for AI beneficiaries using the same metrics: revenue growth, P/E ratios, and maybe gross margins. Here's what these screens miss entirely:
- Broadcom (AVGO) has 40% of revenue concentrated in 5 customers—a fact buried in Risk Factors
- NVIDIA (NVDA) depends on a single "Customer A" for 22% of total revenue
- Arista Networks (ANET) derives 35% of revenue from just Microsoft and Meta
- Vertiv (VRT) explicitly pursues diversification to avoid this exact risk
Standard screeners show none of this. They'll tell you all these stocks are "AI plays" without distinguishing concentration risk from diversified quality.
This framework uses 2 signals derived from SEC filing analysis to stress-test AI infrastructure beneficiaries:
- Customer Concentration: The RISK dimension—how vulnerable is the company to a single customer reducing orders?
- Margin Momentum: The QUALITY dimension—is the company's profitability improving or deteriorating?
Together, these signals answer the core question: Is this AI stock worth the risk?
What Is the 2-Signal Framework for AI Beneficiary Analysis?
| Signal | What It Measures | Thresholds | Data Source |
|---|---|---|---|
| Customer Concentration | Revenue dependence on top customers | <15% = LOW, 15-25% = MODERATE, 25-40% = ELEVATED, >40% = HIGH | SEC 10-K/10-Q Risk Factors |
| Margin Momentum (8Q) | Profitability trajectory over 2 years | Positive = Improving, Negative = Declining | Trailing 8-quarter trend calculation |
Why these signals matter together:
- High concentration + improving margins = The company is extracting value despite customer risk
- High concentration + declining margins = Binary risk with no quality offset
- Low concentration + improving margins = Best risk-adjusted profile
- Low concentration + declining margins = Structural problem, not customer-related
What Is Customer Concentration Risk and Why Does It Matter?
Customer concentration creates binary risk—when 2-3 customers represent a significant portion of revenue, losing one isn't a headwind, it's potentially catastrophic.
SEC rules require disclosure when any single customer exceeds 10% of revenue. Companies often disclose aggregate concentration (e.g., "top 5 customers") in the Risk Factors section of 10-K and 10-Q filings.
Concentration Comparison
| Ticker | Concentration | Top Exposure | Risk Level | SEC Quote |
|---|---|---|---|---|
| AVGO | 40% | Top 5 customers | HIGH | "top five end customers accounted for approximately 40% of our net revenue" |
| ANET | 35% | Microsoft + Meta | ELEVATED | "Microsoft represented 20%...Meta Platforms represented approximately 15%" |
| NVDA | 22% | Customer A (single) | ELEVATED | "Customer A alone accounting for 22% of our total revenue" |
| VRT | None | Diversified | LOW | "diversification strategy to avoid over concentration" |
Source: Company 10-Q/10-K filings via MetricDuck Filing Intelligence extraction.
Understanding the Risk
Broadcom (AVGO) has the highest concentration at 40% across its top 5 customers. The company also disclosed that "sales to distributors accounted for 48% of our net revenue," indicating significant channel concentration. However, the VMware acquisition has diversified AVGO's revenue streams beyond semiconductor customers.
Arista Networks (ANET) presents the most binary risk profile. Per their 10-K:
"Sales to Microsoft represented 20%, 18% and 16% of our total revenue for the years ended December 31, 2024, 2023 and 2022, respectively."
The trend shows increasing Microsoft dependence (16% → 20% over 3 years). Combined with Meta's 15%, ANET derives 35% of revenue from just two hyperscaler customers. If either Microsoft or Meta in-houses networking infrastructure or reduces AI capex, ANET's revenue would face a material decline.
NVIDIA (NVDA) disclosed exceptional concentration:
"We generate a significant amount of our revenue from a limited number of indirect customers, some individually representing 10% or more of our revenue."
Specifically, NVDA reported 4 customers exceeding 10% of revenue, with "Customer A" alone at 22%. This is almost certainly a hyperscaler (likely Microsoft, Meta, or Amazon). The binary risk: if any single hyperscaler pauses or reduces AI infrastructure orders, NVDA's revenue impact would be immediate and material.
Vertiv (VRT) explicitly pursues the opposite strategy:
"diversification strategy to avoid over concentration"
VRT does not disclose any customer above 10%—the SEC disclosure threshold. This implies no single customer exceeds 10% of revenue, the lowest concentration risk among the four companies analyzed.
Product Link: View ANET Filing Intelligence | View NVDA Filing Intelligence
What Does 8-Quarter Margin Momentum Reveal About Quality?
Point-in-time metrics miss trajectory. A company with 30% margins declining is a worse investment than one at 20% margins improving.
Our 8-quarter trend analysis (Q.TREND8) captures the direction of profitability over approximately 2 years—long enough to filter noise, short enough to reflect current dynamics.
ROIC Trends (Return on Invested Capital)
ROIC measures how efficiently the company generates profit from all invested capital. Higher ROIC indicates a competitive moat.
| Ticker | Current ROIC | ROIC 8Q Ago | 8Q Change | Interpretation |
|---|---|---|---|---|
| NVDA | 204.4% | 184.0% | +20.38pp | Exceptional, improving |
| ANET | 160.8% | 106.3% | +54.54pp | Exceptional, accelerating |
| AVGO | 72.7% | 17.8% | +54.91pp | Strong, accelerating (VMware) |
| VRT | 35.7% | 34.0% | +1.65pp | Moderate, stable |
Key Finding: AVGO and ANET show the strongest ROIC momentum, both improving by ~55 percentage points over 8 quarters. NVDA's 204% ROIC is exceptional in absolute terms, but the pace of improvement has moderated.
Gross Margin Trends
Gross margin measures pricing power—the spread between revenue and cost of goods sold. Improving gross margins indicate either pricing strength or cost efficiency.
| Ticker | Current GM | GM 8Q Ago | 8Q Change | Interpretation |
|---|---|---|---|---|
| AVGO | 67.1% | 60.8% | +6.33pp | Strong improvement (VMware) |
| VRT | 37.8% | 36.7% | +1.14pp | Steady improvement |
| ANET | 64.6% | 64.9% | -0.31pp | Flat |
| NVDA | 73.4% | 76.0% | -2.56pp | Slight decline |
Key Finding: AVGO shows the strongest gross margin improvement at +6.33pp, driven by VMware integration synergies. Interestingly, NVDA's gross margin is declining slightly despite exceptional ROIC—suggesting the AI compute dominance isn't translating to improving unit economics.
The Divergence Story
NVDA and ANET demonstrate a notable pattern: improving ROIC but flat-to-declining gross margins. This suggests:
- Revenue growth is outpacing cost inflation (scaling benefits)
- Capital efficiency is improving (better asset utilization)
- BUT unit-level pricing power may be weakening
For investors, this divergence matters. Companies with improving both ROIC and gross margins (like AVGO) demonstrate quality across multiple dimensions. Companies with improving ROIC but declining margins may face future pressure as growth moderates.
How Does Each AI Beneficiary Stack Up?
Broadcom (AVGO): High Risk, High Reward
Concentration: 40% (top 5 customers) | Margin Momentum: +6.33pp GM | ROIC: 72.7% (+54.91pp)
Broadcom presents a complex risk/reward profile. The 40% customer concentration is the highest in our analysis, yet the company shows the strongest margin improvement trajectory.
The VMware Integration Story:
AVGO's margin expansion is primarily driven by the November 2023 VMware acquisition. The integration has been aggressive:
- Gross margin improved from 60.8% to 67.1% (+6.33pp)
- ROIC accelerated from 17.8% to 72.7% (+54.91pp)
- Operating leverage from software revenue mix shift
The VMware acquisition also diversifies AVGO's revenue base. Pre-acquisition, AVGO was heavily dependent on semiconductor customers (hyperscalers, OEMs). Post-acquisition, software and services now represent a meaningful portion of revenue, partially offsetting semiconductor concentration risk.
The Risk:
AVGO's 40% top-5 concentration and 48% distributor concentration remain material risks. If hyperscaler AI capex slows or a major distributor relationship ends, revenue impact would be significant. The company's investor summary notes:
"top five end customers accounted for approximately 40% of our net revenue"
Verdict: QUALITY + RISK. The VMware synergies demonstrate execution capability, but concentration remains elevated. Suitable for investors comfortable with binary customer risk in exchange for exceptional margin momentum.
Product Link: AVGO Filing Intelligence | AVGO Peer Comparison
NVIDIA (NVDA): AI Dominance with Concentration Trade-offs
Concentration: 22% (single customer) | Margin Momentum: -2.56pp GM | ROIC: 204.4% (+20.38pp)
NVIDIA is the undisputed leader in AI compute infrastructure, with a 204.4% ROIC that towers above peers. But this dominance comes with hidden concentration risk.
The Customer A Problem:
NVDA disclosed that four customers individually exceed 10% of revenue, with "Customer A" alone at 22%. The company's risk factors note:
"We generate a significant amount of our revenue from a limited number of indirect customers, some individually representing 10% or more of our revenue."
This 22% single-customer concentration is unusual for a company of NVDA's scale. Most large-cap technology companies have no single customer exceeding 5-10%. The identity is not disclosed, but hyperscaler AI infrastructure buyers (Microsoft, Meta, Amazon, Google) are the likely candidates.
The Margin Pressure Signal:
Despite exceptional ROIC, NVDA's gross margin is declining (-2.56pp over 8 quarters). This could indicate:
- Product mix shift toward lower-margin data center hardware
- Competitive pressure on AI chip pricing
- Customer negotiating leverage as concentration increases
The combination of 22% single-customer exposure and declining gross margins suggests NVDA's moat, while formidable, may face pressure as customers gain bargaining power.
The Quality Offset:
NVDA's 204% ROIC is exceptional by any standard. The +20.38pp improvement over 8 quarters demonstrates the company is efficiently deploying capital despite margin headwinds. For now, volume growth is more than offsetting unit economics pressure.
Verdict: ELEVATED RISK, EXCEPTIONAL QUALITY. NVDA's AI dominance is undeniable, but the 22% single-customer concentration and declining gross margins are warning signs. Investors should monitor whether margin pressure accelerates as AI infrastructure spending cycles evolve.
Product Link: NVDA Filing Intelligence | NVDA Peer Comparison
Arista Networks (ANET): Hyperscaler Binary Risk
Concentration: 35% (Microsoft + Meta) | Margin Momentum: -0.31pp GM | ROIC: 160.8% (+54.54pp)
Arista Networks presents the most binary risk profile among the four companies. With Microsoft (20%) and Meta (15%) comprising 35% of revenue, ANET's fortunes are tied directly to hyperscaler data center spending.
The Hyperscaler Dependence:
ANET's 10-K explicitly discloses the concentration:
"Sales to Microsoft represented 20%, 18% and 16% of our total revenue for the years ended December 31, 2024, 2023 and 2022, respectively."
The trend is notable: Microsoft dependence increased from 16% to 20% over three years. Combined with Meta's 15%, ANET derives 35% of revenue from just two customers. This is the definition of binary risk—if either hyperscaler pauses data center builds or in-houses networking, ANET faces material revenue decline.
The Quality Story:
Despite concentration risk, ANET's fundamentals are exceptional:
- 160.8% ROIC (second only to NVDA)
- ROIC improved +54.54pp over 8 quarters
- Gross margin stable at 64.6% (-0.31pp is essentially flat)
The company is clearly generating exceptional returns on capital. The question is whether this quality persists if hyperscaler spending moderates.
The AI Infrastructure Play:
ANET is a direct beneficiary of AI data center buildout. The company's networking solutions are essential for connecting GPU clusters in AI training environments. As long as hyperscaler AI capex continues, ANET's revenue tailwind persists.
Verdict: HIGH RISK, EXCEPTIONAL ROIC. ANET is the highest-quality binary bet on hyperscaler AI infrastructure spending. The 35% Microsoft+Meta concentration is a feature and a bug—it provides AI exposure but creates existential risk if either customer reduces orders. Suitable for investors with high conviction on continued hyperscaler capex.
Product Link: ANET Filing Intelligence | ANET Peer Comparison
Vertiv (VRT): The Low-Risk Contrast
Concentration: None disclosed | Margin Momentum: +1.14pp GM | ROIC: 35.7% (+1.65pp)
Vertiv is the outlier in this analysis—deliberately so. The company explicitly pursues diversification to avoid the concentration risks that characterize AVGO, NVDA, and ANET.
The Diversification Strategy:
VRT's SEC filings explicitly state:
"diversification strategy to avoid over concentration"
The company does not disclose any customer above the 10% SEC threshold, implying no single customer accounts for more than 10% of revenue. This is the lowest concentration risk among AI infrastructure beneficiaries.
The AI Infrastructure Angle:
VRT provides power and thermal management solutions for data centers—essential infrastructure for AI compute. As GPU clusters generate significant heat and require substantial power, VRT benefits from AI buildout without the customer concentration of chip and networking vendors.
The business model advantages:
- Sells to diverse data center operators, not just hyperscalers
- Provides maintenance and services revenue streams
- Less exposed to any single customer's capex cycle
The Quality Profile:
VRT's metrics are solid but not exceptional:
- 35.7% ROIC (lowest among the four)
- ROIC improved +1.65pp over 8 quarters (steady)
- Gross margin improved +1.14pp (steady)
The trajectory is positive but modest. VRT trades some upside potential for lower concentration risk.
Verdict: LOW RISK, QUALITY PROFILE. VRT offers AI infrastructure exposure without hyperscaler concentration risk. The trade-off is more modest ROIC and margin momentum compared to NVDA, ANET, and AVGO. Suitable for investors seeking AI exposure with a more defensive risk profile.
Product Link: VRT Filing Intelligence | VRT Peer Comparison
What Is the Risk/Reward Profile for Each Stock?
| Ticker | Concentration | Margin Momentum | ROIC | Risk Level | Quality | Verdict |
|---|---|---|---|---|---|---|
| AVGO | 40% (HIGH) | +6.33pp (STRONG) | 72.7% | HIGH | HIGH | High Risk, High Reward |
| ANET | 35% (ELEVATED) | -0.31pp (FLAT) | 160.8% | HIGH | HIGH | Binary Hyperscaler Bet |
| NVDA | 22% (ELEVATED) | -2.56pp (DECLINING) | 204.4% | ELEVATED | EXCEPTIONAL | Dominant but Concentrated |
| VRT | None (LOW) | +1.14pp (STEADY) | 35.7% | LOW | MODERATE | Defensive AI Play |
Source: Q3 2025 10-Q filings via MetricDuck Filing Intelligence extraction and 8-quarter trend calculation.
How Do I Apply This Framework to Other Stocks?
Step 1: Check Customer Concentration
Pull the company's latest 10-K or 10-Q and search the Risk Factors section for:
- "customer concentration"
- "significant customer"
- "represented X% of revenue"
Apply our thresholds:
- <15% = LOW risk
- 15-25% = MODERATE risk
- 25-40% = ELEVATED risk
- >40% = HIGH risk
Step 2: Calculate Margin Momentum
Compare current margins to 8 quarters ago:
- Gross Margin: Measures pricing power
- ROIC: Measures overall capital efficiency
Positive trend = Quality signal Negative trend = Warning sign
Step 3: Combine for Risk/Reward Assessment
| Concentration | Margin Trend | Profile |
|---|---|---|
| LOW | Improving | Best risk-adjusted |
| LOW | Declining | Structural concern |
| HIGH | Improving | High risk, high reward |
| HIGH | Declining | Avoid |
Screen These Stocks Yourself
Use MetricDuck's tools to apply this framework:
Filing Intelligence for Concentration Analysis:
- AVGO Filing Intelligence - 40% top-5 concentration
- NVDA Filing Intelligence - 22% single customer
- ANET Filing Intelligence - 35% hyperscaler concentration
- VRT Filing Intelligence - Diversification strategy
Screen for Similar Profiles:
- Screen by ROIC + Margin Trends - Find quality profiles
- Custom Peer Groups - Build your own comparisons
Get Notified When Q4 2025 Data Arrives
This analysis uses Q3 2025 10-Q filings. When Q4 2025 annual reports (10-K) are filed, our Filing Intelligence will automatically extract updated concentration and margin data.
Subscribe to filing alerts to receive updated analysis:
| Company | Alert Link |
|---|---|
| AVGO - Broadcom | Subscribe to AVGO alerts |
| NVDA - NVIDIA | Subscribe to NVDA alerts |
| ANET - Arista Networks | Subscribe to ANET alerts |
| VRT - Vertiv | Subscribe to VRT alerts |
Or manage all your filing alerts from your dashboard.
Methodology Notes
Data Sources:
- All metrics calculated from SEC EDGAR XBRL filings
- Customer concentration extracted from 10-K/10-Q Risk Factors sections using Filing Intelligence
- 8-quarter trends use trailing twelve months (TTM) calculations
Metric Definitions:
- Customer Concentration: Percentage of revenue from disclosed top customers (SEC requires disclosure for >10%)
- Gross Margin: (Revenue - Cost of Revenue) / Revenue
- ROIC: Net Operating Profit After Tax / Invested Capital
- 8-Quarter Trend: Change from 8 quarters ago to current period
Limitations:
- Customer concentration is only disclosed when >10% threshold is met; companies below this threshold may still have meaningful concentration
- Companies may use different aggregation methods (top 5 vs. single customer vs. end customer vs. distributor)
- Hyperscaler identities are often undisclosed ("Customer A"); our attributions are based on public reporting and inference
- ROIC calculations use TTM operating income and may include one-time items; AVGO's ROIC reflects VMware integration
- 8-quarter trends may span fiscal year boundary changes or accounting policy updates
- Margin momentum can be distorted by acquisition accounting (e.g., AVGO's VMware purchase price allocation)
External References:
- SEC EDGAR — Primary source for 10-K/10-Q filings
- Investopedia: Customer Concentration — Definition and risk framework
- CFA Institute: Return on Invested Capital — ROIC methodology and interpretation
- Investopedia: Gross Margin — Gross margin calculation and analysis
Next Update: After Q4 2025 earnings (late January 2026)
Disclaimer
This analysis is for educational and informational purposes only. It does not constitute investment advice, and you should not rely on it as such.
Important considerations:
- Past performance does not guarantee future results
- Customer concentration can change rapidly based on contract renewals
- AI infrastructure spending is cyclical and subject to economic conditions
- The author holds no positions in any of the stocks mentioned
- Always consult a qualified financial advisor before making investment decisions
- Data is sourced from SEC filings, which may contain errors or be subject to restatement
Conflict of Interest Disclosure: MetricDuck provides financial analysis tools. This content demonstrates our analytical capabilities but should not be construed as a recommendation to buy or sell any security.
Related Analysis
- AI Data Center Risk Screen: 4 Signals Wall Street Misses
- ROIC Stock Screening: 5-Step Framework + Sector Benchmarks
- Cash Flow Quality Analysis Framework
Last updated: December 22, 2025. Next update scheduled after Q4 2025 earnings (late January 2026).
Explore More AI Infrastructure Research
This article is part of our comprehensive AI Infrastructure Hub, which covers capex efficiency metrics, depreciation manipulation detection, and supply chain analysis.
Related AI infrastructure research:
- How to Track AI Capex Efficiency — 3-metric framework for monitoring hyperscaler spending
- AI Data Center Risk Screen — 4 signals Wall Street misses
- Magnificent 7 Valuation Scorecard — Sector-adjusted grades for hyperscalers
MetricDuck Research
SEC filing analysis and XBRL data extraction