AI Stocks With Lowest Risk, Highest Upside

The Problem

We don’t know if TPU or GPU is going to win. So no betting on GOOG nor NVDA for us. But we know compute in general is a secular trend and want to ride it.

The Lowest Risk: Memory, advanced packaging, and high‑speed networking suppliers appear structurally better positioned, as their demand is driven by aggregate compute growth regardless of whether workloads originate from chatbots, recommenders or automotive, warehouse automations.

From this perspective, four companies have emerged as critical infrastructure providers: Vertiv (VRT), Credo Technology (CRDO), Astera Labs (ALAB), and Arista Networks (ANET). While all benefit from AI tailwinds, they occupy different positions in the value chain with varying moat strengths, customer criticality, and valuations. This analysis examines which companies merit premium valuations and which represent investment opportunities.

Part I: Understanding the Moats

Vertiv (VRT): The Physical Infrastructure Gatekeeper

Moat Strength: ★★★★☆ (4/5)

Vertiv's competitive advantages are multi-layered:

Technical Moats:

• Thermal Management Expertise: Liquid cooling for 140kW AI racks requires years of R&D and deep physics understanding. Vertiv's collaboration with CoreWeave on NVIDIA GB300 NVL72 deployment demonstrates first-mover advantage in next-generation cooling

• System Integration Complexity: Gray and white space convergence creates interdependencies between power distribution, cooling, racks, and monitoring that favor integrated providers

• Installed Base: Existing relationships with hyperscalers and service cycle revenue create switching costs

Scale Advantages:

• Manufacturing footprint across U.S. and Europe (expanded through Great Lakes acquisition)

• Ability to deliver prefabricated solutions (SmartRun platform) reduces deployment time from months to weeks

• Global supply chain optimization to navigate tariff environment

Weaknesses:

• Capital-intensive business model limits ROIC expansion

• Faces competition from electrical equipment giants (Schneider Electric, Eaton) and new entrants

• Exposed to cyclical capex spending patterns

Verdict: Vertiv has a solid moat built on technical complexity and customer relationships, but it's not impregnable. The transition to liquid cooling creates a temporary window where expertise matters more than scale.

Credo Technology (CRDO): The Connectivity Specialist

Moat Strength: ★★★☆☆ (3/5)

Credo occupies a critical but competitive position:

Technology Moats:

• Active Electrical Cables (AEC): Proprietary technology offering lower power consumption than optical alternatives for short-reach connections

• SerDes IP Portfolio: Digital signal processor expertise across multiple protocols (Ethernet, PCIe)

• Energy Efficiency Leadership: Critical advantage as data center power costs escalate

Commercial Moats:

• Design-In Cycles: 18-24 month qualification periods create temporary stickiness once designed into platforms

• Hyperscaler Relationships: Deep engagement with top-3 cloud providers (60%+ of revenue)

Significant Weaknesses:

• Intense Competition: Battles Broadcom, Marvell, and Astera Labs across multiple product lines

• Patent Licensing: Recent Siemon agreement suggests defensive positioning

• Customer Concentration Risk: Top 3 customers >60% of revenue; single platform loss could be devastating

• Limited Pricing Power: As connectivity becomes commoditized, margins face pressure

Verdict: Credo has technology advantages but operates in a battlefield with well-funded competitors. The moat is real but narrow, requiring constant innovation to maintain position.

Astera Labs (ALAB): The Protocol Translation Leader

Moat Strength: ★★★★☆ (4/5)

Astera has carved out a defensible niche:

Technical Moats:

• PCIe 6.0 Retimer Leadership: First-to-market advantage in critical signal integrity components

• CXL 3.0 Expertise: Positioned at the intersection of compute and memory disaggregation

• Scorpio Fabric Switches: Moving up the value chain into rack-scale orchestration

• Software-Defined Architecture: Cosmos platform creates differentiation beyond hardware

Strategic Moats:

• Design Win Momentum: Present in NVIDIA, AMD, Intel platforms creates embedded position

• Partnerships: Alchip collaboration for custom silicon strengthens ecosystem

• Market Timing: Arriving as PCIe Gen 6 and CXL become mandatory for AI servers

Vulnerabilities:

• Platform Risk: Custom solutions from hyperscalers (NVIDIA NVLink, custom ASICs) could bypass merchant silicon

• ESUN Consortium Threat: October 2025 announcement of industry collaboration on connectivity could marginalize independent players

• Execution Requirements: Must maintain technology leadership across 4 product families simultaneously

Verdict: Astera has strong moats based on protocol expertise and first-mover advantages, but faces risks from vertical integration trends. The moat is deep but could be circumvented by sufficiently motivated customers.

Arista Networks (ANET): The Networking Standard-Bearer

Moat Strength: ★★★★★ (5/5)

Arista possesses the strongest competitive position:

Software Moats:

• EOS Platform: CloudVision management creates multi-year lock-in and recurring revenue

• Network Effect: As more customers deploy EOS, third-party integrations proliferate

• Operational Simplicity: Zero-touch provisioning and automation reduce OpEx, creating 3-5 year switching costs

Technical Moats:

• AI Networking Leadership: 800G Etherlink portfolio competing directly with NVIDIA's InfiniBand

• Full-Stack Integration: Hardware, software, and management unified under single vendor

• Protocol Flexibility: Support for Ethernet, InfiniBand, and emerging Ultra Ethernet standards

Strategic Moats:

• Hyperscaler Penetration: Microsoft (20% of revenue), Meta (15%) create reference architectures

• Market Share Gains: Taking share from Cisco in both cloud and campus markets

• Gross Margin Profile: 60%+ gross margins indicate pricing power

Limited Weaknesses:

• Reliance on Broadcom for merchant silicon creates supply risk

• Competition from NVIDIA (spectrum switches) and white-box alternatives

• Campus market growth slower than cloud networking

Verdict: Arista has the most durable moat in the group, built on software differentiation, customer lock-in, and operational excellence. This is reflected in its ability to sustain premium margins despite intense competition.

Part II: Essential vs. Discretionary Analysis

Mission-Critical Assessment Framework

To evaluate how essential each company's products are to customer operations, we analyze:

1. Substitutability: Can customers easily switch to alternatives?

2. Impact of Failure: What happens if the component fails?

3. Integration Depth: How embedded is the solution in customer infrastructure?

4. Budget Priority: Where does this fall in capex allocation during downturns?

Vertiv: Mission-Critical Physical Layer

Essential Rating: ★★★★★ (5/5)

Why It's Essential:

• Data centers cannot operate without power and cooling infrastructure - this is literally "keep the lights on" spending

• AI workloads generating 140kW+ per rack make liquid cooling non-negotiable, not optional

• Infrastructure refresh cycles align with IT equipment cycles (3-5 years), creating predictable demand

Customer Perspective: When a hyperscaler plans a $10 billion data center, Vertiv's infrastructure represents ~15-20% of total capex. While large in absolute terms, it's:

• Non-Deferrable: Without power/cooling, servers are paperweights

• Insurance Spending: Prevents catastrophic failures that could cost millions per hour

• Required for Compliance: Meets tier-3/tier-4 data center certifications

Recession Behavior: Vertiv's backlog of $8.5B (up 21% YoY) demonstrates forward visibility. Even if hyperscalers slow new data center builds, existing facilities require continuous infrastructure investment. The transition to liquid cooling creates a one-time upgrade cycle independent of broader IT spending.

Conclusion: Vertiv is about as essential as infrastructure spending gets. The only scenario where spending declines is a complete halt to AI investment - unlikely given committed capex.

Credo: Necessary Connectivity, But Commoditizing

Essential Rating: ★★★☆☆ (3/5)

Why It's Necessary:

• High-speed connectivity is mandatory for AI clusters - can't train models without moving data

• Energy efficiency matters as power costs approach 40% of TCO

• Active electrical cables offer 30-40% cost savings vs. optical for certain distances

Why It's Vulnerable:

• Multiple Suppliers: Customers can source from Credo, Broadcom, Marvell, or internal solutions

• Price Sensitivity: At $300-500 per cable, large deployments create incentives to negotiate aggressively or vertically integrate

• Technology Substitution: Optical connectivity improving in cost/watt, eroding AEC advantages

Customer Perspective: A hyperscaler building a 32,000 GPU cluster might spend $50-100M on connectivity components. While significant, this represents:

• ~5% of GPU costs: Dwarfed by accelerator spending

• Negotiable Spending: Multiple vendors enable competitive procurement

• Optimization Target: Engineering teams seek to reduce per-unit costs annually

Recession Behavior: In a downturn, customers would:

1. Delay next-generation upgrades (stick with Gen 5 instead of Gen 6)

2. Negotiate harder on pricing

3. Explore vertical integration for high-volume components

Credo's Q1 FY2026 revenue of $223M (up 274% YoY) reflects current AI boom, but this growth rate is unsustainable if spending moderates.

Conclusion: Credo provides necessary functionality, but lacks the pricing power and customer lock-in of mission-critical providers. It's essential for today's AI infrastructure but faces substitution risk.

Astera Labs: Critical Protocol Translation

Essential Rating: ★★★★☆ (4/5)

Why It's Critical:

• PCIe Retimers: Non-optional for signal integrity in high-speed server designs. Without retimers, PCIe Gen 5/6 cannot reliably transmit over backplane distances

• CXL Controllers: Enabling memory disaggregation - a architectural shift, not incremental improvement

• Scorpio Switches: Rack-scale orchestration becoming table-stakes for AI efficiency

Why It's Not 5-Stars:

• Hyperscaler Alternatives: Large cloud providers developing custom silicon (Google TPU interconnect, AWS Trainium networking)

• ESUN Threat: Industry consortium could standardize open alternatives

• Platform Dependency: If NVIDIA or AMD change architectures, Astera must adapt or lose relevance

Customer Perspective: For a 100,000 GPU AI factory:

• Astera components represent ~2-3% of total system cost

• But enable ~10-15% performance improvements through better signal integrity and memory pooling

• ROI calculation: $30M Astera spending enables $200M+ additional utilization value

Recession Behavior: Astera benefits from multiple drivers:

1. Technology Transitions: PCIe Gen 6 and CXL 3.0 adoption continues regardless of spending levels

2. Retrofit Market: Existing data centers upgrading connectivity creates demand independent of new builds

3. Diversification: Moving into switches reduces dependency on single product line

However, if AI spending craters, Astera's 75%+ gross margins could compress as customers demand pricing concessions.

Conclusion: Astera occupies a critical position in the AI infrastructure stack with strong technical moats, but faces existential risk from vertical integration. Essential for now, but durability depends on maintaining innovation leadership.

Arista: The Networking Backbone

Essential Rating: ★★★★★ (5/5)

Why It's Essential:

• Network is the Platform: AI training requires lossless, low-latency fabric - networking is as critical as GPUs

• Operational Lock-In: CloudVision management embedded in customer workflows; switching costs measured in years

• Back-End AND Front-End: Arista serves both AI training clusters (back-end) and campus/edge (front-end)

Customer Validation:

• Microsoft: 20% of Arista's revenue, multi-year commitment

• Meta: 15% of Arista's revenue, betting on Ethernet for AI

• These aren't trials - these are production deployments at planetary scale

Why It's Truly Mission-Critical: Unlike components that can be swapped, network architecture decisions have 5-10 year implications:

• Training Job Efficiency: Network bottlenecks waste billions in GPU cycles

• Service Reliability: Customer-facing applications depend on network uptime

• TCO Impact: Arista's automation reduces operational costs by 30-50%

Recession Behavior: Historical data shows Arista's resilience:

• 2022-2023: During tech layoffs and capex pullback, Arista maintained growth

• Reason: Cloud titans don't stop expanding networks; they optimize efficiency

• AI transition: Even if growth slows, migration from InfiniBand to Ethernet creates multi-year tailwind

Conclusion: Arista is as essential as networking infrastructure gets. Customers are locked in operationally, not just technically. This explains sustained 60%+ gross margins and pricing power.

Part III: Valuation Analysis (Reverse DCF)

Methodology: What's Already Priced In?

Rather than forecast the future, we'll work backward: Given current stock prices, what growth and margin assumptions are embedded? This reveals whether expectations are reasonable or fantastical.

Framework:

• Discount Rate: 10% (appropriate for high-growth tech)

• Terminal Growth: 3% (GDP growth)

• Margin Targets: Based on industry benchmarks

• Time Horizon: 10 years to stabilization

Vertiv (VRT): Fairly Valued for Infrastructure Compounder

Current Metrics (Q2 2025):

• Market Cap: ~$40B (est. based on Q2 presentation)

• Revenue (TTM): ~$10B (run-rate from Q2: $2.6B × 4)

• Operating Margin: 18.5% (Q2 2025)

• EV/Sales: ~4.0x

• Growth Guidance: +22% FY2025E

Reverse DCF Analysis:

What's Priced In:

• Revenue CAGR: 15-18% through 2030

• Operating Margin: Expansion to 22-23% (from 18.5%)

• Terminal Value: Stabilized 10% FCF yield

Is This Achievable? ✅ Growth Case:

• AI data center buildout: Multi-year cycle (2024-2030)

• Liquid cooling transition: Capturing share of $50B+ TAM

• Great Lakes acquisition: Adds ~$200M revenue at higher margins

✅ Margin Case:

• Americas already at 24% operating margin

• Scale economies: $10B+ revenue enables fixed cost leverage

• Pricing power: Essential infrastructure during capacity constraints

• Target: 22-23% operating margin by 2027 is realistic

Risks:

• Tariff impact: Could pressure near-term margins by 50-100 bps

• Cyclicality: Infrastructure spending follows IT capex with 6-12 month lag

• Competition: Schneider, Eaton have deep pockets

Valuation Verdict: FAIRLY VALUED

At 4x EV/Sales for a business growing 15-20% with 20%+ operating margins, Vertiv trades at fair value. Not a screaming buy, but reasonable risk/reward for infrastructure exposure.

Target Multiple: 4-5x EV/Sales justifiable given:

• Mission-critical positioning

• Multi-year visibility (backlog)

• Margin expansion runway

Credo Technology (CRDO): Extraordinarily Rich Valuation

Current Metrics (Q2 FY2026 est.):

• Market Cap: ~$28B

• Revenue (LTM): ~$600M

• Operating Margin: 27% (non-GAAP Q2 FY2026E)

• EV/Sales: ~47x

• Forward P/E: ~98x

• Growth: 274% YoY Q1 FY2026

Reverse DCF Analysis:

What's Priced In:

• Revenue Growth: 60%+ CAGR for 5 years, then 30%+ thereafter

• Operating Margin: Sustained 40%+ (best-in-class)

• Market Cap: Must reach $600B+ by 2035 to justify current price

• Implied TAM Capture: Dominant market share in $25B+ connectivity market

Reality Check:

Growth Requirements:

• 2025E: $600M → 2030E: ~$7-8B revenue to justify valuation

• This implies: Credo becomes larger than current Broadcom connectivity division

• Problem: Multiple competitors (Broadcom, Marvell, Astera, Nvidia) targeting same customers

Margin Challenges:

• Current 67% gross margin, 27% operating margin is exceptional

• But: As volumes scale, pricing pressure intensifies

• Hyperscalers: Notorious for squeezing suppliers once past design-in phase

• Reality: Operating margin likely peaks at 30-35%, not 40%+

Is This Achievable? ❌ Highly Unlikely

Why Current Valuation is Untenable:

1. Competition: Cannot sustain dominance with Broadcom/Marvell/Nvidia all competing

2. Customer Leverage: Hyperscaler concentration gives buyers pricing power

3. Commoditization: Connectivity becoming standardized reduces differentiation

4. Insider Selling: $150M sold in 3 months suggests management skepticism

Comparable Analysis:

• Broadcom: EV/Sales ~15x, with diversified portfolio and entrenched positions

• Marvell: EV/Sales ~12x, with similar connectivity exposure

• Credo: EV/Sales ~47x, with concentrated customer base

Valuation Verdict: SIGNIFICANTLY OVERVALUED

Credo would need to execute flawlessly for a decade, capture dominant market share, and maintain best-in-class margins to justify 47x EV/Sales. Current price embeds near-zero margin for error.

Fair Value Estimate: 15-20x EV/Sales

• Implies: $5-7B market cap vs. current $28B

• Downside Risk: 75%+ if growth disappoints or competition intensifies

Conclusion: Credo is a good business at an extraordinary price. The risk/reward is heavily skewed to downside.

Astera Labs (ALAB): Expensive But Defensible

Current Metrics (Q3 2025):

• Market Cap: ~$28B

• Revenue (LTM): ~$600M

• Operating Margin: 6-7% GAAP, ~40% non-GAAP (Q3 2025E)

• EV/Sales: ~46x (similar to Credo)

• Forward P/E: ~98x

• Growth: 150%+ YoY

Reverse DCF Analysis:

What's Priced In:

• Revenue Growth: 50%+ CAGR through 2028, then 30%+ through 2032

• Operating Margin: Expansion to 40%+ (non-GAAP) sustained

• Market Leadership: Dominant position in PCIe retimers, CXL, and switches

• TAM Expansion: $12B+ addressable market by 2028

Is This Achievable? ⚠️ Possible But Challenging

Growth Case: ✅ Strong Fundamentals:

• Technology Leadership: First to market with PCIe Gen 6, CXL 3.0

• Multiple Vectors: 4 product families (Aries, Taurus, Leo, Scorpio) diversify risk

• Design Win Momentum: Embedded in NVIDIA/AMD platforms

• Margin Profile: 75%+ gross margins provide cushion

✅ Market Drivers:

• PCIe Gen 6 Transition: Mandatory upgrade cycle (2025-2027)

• CXL Adoption: Memory disaggregation becoming architectural requirement

• Scorpio Switches: 10%+ revenue in 2025, doubling in 2026

❌ Significant Risks:

• ESUN Consortium: October 2025 announcement of open networking standards

• Vertical Integration: Hyperscalers developing custom connectivity silicon

• Competition: Broadcom, Marvell have more resources and customer relationships

• Customer Concentration: Top customers drive 60%+ of revenue

Margin Analysis:

• Current: 76% gross, 41.7% non-GAAP operating (Q3 2025)

• Best-Case: Sustain 70%+ gross, 38-40% operating

• Bear-Case: Gross margin compresses to 60%, operating to 25% as competition intensifies

• Realistic: 65-70% gross, 32-35% operating by 2028

Valuation Scenarios:

Bull Case (40% probability):

• Maintains technology lead, Scorpio succeeds, CXL becomes standard

• 2030 Revenue: $3-4B, Operating Margin: 38%

• Justified Multiple: 20-25x EV/Sales

• Implied 2030 Market Cap: $60-100B

• IRR from Today: 8-12% (underwhelming given risk)

Base Case (45% probability):

• Strong execution but faces competition, margins compress modestly

• 2030 Revenue: $2-3B, Operating Margin: 32%

• Justified Multiple: 12-15x EV/Sales

• Implied 2030 Market Cap: $25-45B

• IRR from Today: -1% to +4% (flat to slightly down)

Bear Case (15% probability):

• ESUN succeeds, hyperscalers integrate vertically, pricing pressure

• 2030 Revenue: $1.5B, Operating Margin: 22%

• Justified Multiple: 8-10x EV/Sales

• Implied 2030 Market Cap: $12-15B

• IRR from Today: -10% to -8% annualized (significant loss)

Valuation Verdict: FULLY VALUED TO OVERVALUED

Astera has better fundamentals than Credo (diversified products, technology leadership, margin profile), but current valuation requires perfect execution. At 46x EV/Sales, there's limited upside even in bull case.

Fair Value Estimate: 20-30x EV/Sales

• Implies: $12-18B market cap vs. current $28B

• Downside Risk: 35-50% if growth slows or competition intensifies

Conclusion: Astera is a high-quality business trading at a full-to-rich valuation. Better than Credo, but still expensive. Would be attractive at $120-140/share (current ~$170).

Arista Networks (ANET): Premium Valuation, Premium Business

Current Metrics (2024):

• Market Cap: ~$100B (based on recent range)

• Revenue (2024): $7.1B (up 17% YoY)

• Operating Margin: 41%+ (best-in-class)

• EV/Sales: ~14x

• Forward P/E: ~35x

• FCF Yield: ~2.5%

Reverse DCF Analysis:

What's Priced In:

• Revenue Growth: 15-18% CAGR through 2030

• Operating Margin: Sustained 40%+ (already achieved)

• Market Share Gains: Continued penetration of cloud and campus

• AI Networking: Winning share from NVIDIA InfiniBand

Is This Achievable? ✅ Highly Likely

Growth Case: ✅ Structural Tailwinds:

• Cloud Networking: Hyperscalers (MSFT, META) locked into multi-year deployments

• AI Transition: Ethernet vs. InfiniBand battle tilting toward Arista

• Campus Modernization: Taking share from Cisco in enterprise market

• Software Revenue: CloudVision subscriptions growing faster than hardware

✅ Competitive Moat:

• Customer Lock-In: EOS platform creates 3-5 year switching costs

• Gross Margins: 60%+ sustained through 3 economic cycles

• Pricing Power: Demonstrated ability to maintain margins despite competition

✅ Financial Profile:

• ROIC: Estimated 25-30% (excellent capital efficiency)

• Cash Generation: $2B+ annual FCF, zero debt

• Capital Allocation: Returning cash via buybacks, M&A optionality

Margin Analysis:

• Current: 60%+ gross, 41%+ operating

• Risk: Modest compression if AI networking becomes commoditized

• Realistic: Gross margins stay 58-62%, operating 38-42%

• Durability: Software mix increasing, improving margin mix

Valuation Scenarios:

Bull Case (30% probability):

• Wins AI networking battle, campus share gains accelerate

• 2030 Revenue: $18-20B, Operating Margin: 43%

• Justified Multiple: 18-20x EV/Sales

• Implied 2030 Market Cap: $320-400B

• IRR from Today: 12-15% (attractive)

Base Case (55% probability):

• Solid execution, maintains market position, steady growth

• 2030 Revenue: $14-16B, Operating Margin: 40%

• Justified Multiple: 14-16x EV/Sales

• Implied 2030 Market Cap: $200-250B

• IRR from Today: 7-9% (reasonable)

Bear Case (15% probability):

• Loses AI share to NVIDIA/Broadcom, campus growth stalls

• 2030 Revenue: $10-12B, Operating Margin: 36%

• Justified Multiple: 10-12x EV/Sales

• Implied 2030 Market Cap: $100-150B

• IRR from Today: 0% to +4% (flat)

Valuation Verdict: FAIRLY VALUED TO MODESTLY OVERVALUED

At 14x EV/Sales, Arista trades at a premium but justified given:

• Quality: Best-in-class margins, ROIC, and customer relationships

• Durability: Software moat provides long-term defensibility

• Visibility: Multi-year hyperscaler commitments reduce uncertainty

Risk/Reward Assessment:

• Upside: 20-30% if AI networking thesis plays out

• Downside: 15-25% if competitive pressures increase

• Base Case: High-single-digit returns with dividend potential

Conclusion: Arista is a high-quality business at a fair-to-full valuation. Not cheap, but justified given moats and execution track record. Suitable for investors seeking quality over deep value.

Part IV: Investment Conclusions

Ranking by Risk/Reward:

1. Vertiv (VRT) - BEST RISK/REWARD ★★★★☆

• Valuation: Fair (4x EV/Sales)

• Moat: Solid (4/5)

• Criticality: Essential (5/5)

• Recommendation: BUY on weakness

• Rationale: Only stock trading near fair value with strong fundamentals. Backlog provides visibility, margins expanding, essential infrastructure spending. Less vulnerable to competition than connectivity plays.

2. Arista Networks (ANET) - HIGH QUALITY AT FAIR PRICE ★★★★☆

• Valuation: Fair to Full (14x EV/Sales)

• Moat: Strongest (5/5)

• Criticality: Essential (5/5)

• Recommendation: HOLD / Accumulate on pullbacks

• Rationale: Premium valuation justified by premium business. Best moat in group, proven management, durable competitive position. Would be strong buy at 10-12x EV/Sales.

3. Astera Labs (ALAB) - QUALITY OVERPRICED ★★★☆☆

• Valuation: Full to Rich (46x EV/Sales)

• Moat: Strong (4/5)

• Criticality: High (4/5)

• Recommendation: HOLD / Consider under $140

• Rationale: Excellent business with technology leadership, but valuation requires flawless execution. Better fundamentals than Credo but still expensive. Wait for 30-40% pullback.

4. Credo Technology (CRDO) - AVOID ★★☆☆☆

• Valuation: Extremely Rich (47x EV/Sales)

• Moat: Moderate (3/5)

• Criticality: Necessary but Vulnerable (3/5)

• Recommendation: SELL / AVOID

• Rationale: Good business at absurd price. Would need to become dominant connectivity player AND sustain 40%+ margins for decade to justify valuation. High risk of 50-75% drawdown.

Portfolio Construction Suggestions:

For Growth Investors:

• Core Position: 50% Arista, 30% Vertiv, 20% Astera (wait for pullback)

• Focuses on quality and duration, accepts moderate valuations for superior moats

For Value Investors:

• Wait: Current prices don't offer margin of safety

• Watchlist: Vertiv at 3x EV/Sales, Arista at 10x EV/Sales, Astera at 20x EV/Sales

• Requires patience and discipline

For Contrarians:

• Thesis: If AI spending disappoints in 2026, all four will get hit

• Play: Wait for broad semiconductor selloff, then build positions in Vertiv and Arista

• Avoid: Credo and Astera lack the moats to recover quickly

Final Thoughts: What Matters Most

After analyzing moats, criticality, and valuations, three conclusions emerge:

1. Infrastructure Beats Components Vertiv and Arista operate at the infrastructure layer where moats are deeper and customer relationships stickier. Connectivity components (Credo, Astera) face commoditization pressure despite current technology advantages.

2. Valuation Discipline Matters Even excellent businesses become poor investments at extreme valuations. Credo and Astera may execute flawlessly and still generate negative returns from current prices.

3. AI Spending is Real, But Expectations Are Astronomical All four companies benefit from genuine AI infrastructure buildout. The question isn't whether they'll grow - it's whether they can grow fast enough to justify valuations that assume perfection.

The Best Investment: Wait for market volatility to create opportunities in Vertiv and Arista. These businesses have durability to compound through cycles, while Credo and Astera face existential risks if AI spending moderates.

Remember: In technology infrastructure, it's better to pay a fair price for a great business than a great price for a fair business. The moat matters more than the entry multiple.

Right now I own ANET, while the others are on my watchlist, which you can access in real-time in our community.

Would you like to stay ahead of opportunities like this? Join our community where we share real-time trade alerts and deep-dive analyses of businesses with true competitive advantages. Don't just trade the market - invest in excellence.

Want to receive our trade alerts and detailed analysis in real-time? Join our community of value investors who understand that pricing power is the ultimate competitive advantage. Receive our trade alerts on your phone? Download the app here