Amazon's moat is unique in its diversification — AWS, retail, advertising, and logistics are each individually defensible. AWS alone at 40% margins on $600B+ ARR run-rate is worth $1.5T+ at a reasonable multiple. $100B+ CapEx is the near-term FCF headwind but infrastructure built once compounds for decades. At 24% discount with Medium uncertainty, this is a core long-term holding.

Google's Search + YouTube + GCP + Android creates uniquely reinforcing advantages. The antitrust risk is real but even forced distribution changes leave Google with the best search product and $247B in cash to adapt. GCP margin inflection toward 20%+ is the underappreciated earnings driver. At 15% discount with Medium uncertainty, GOOGL is a core hold.

META is the most controversial position — exceptional business quality at a genuine discount with a massive uncertain capital investment program. DCF analysis shows $850+ intrinsic value in base case, confirming 38% discount is real. The key question is whether $115-135B CapEx generates returns. At 8.6%, appropriately sized for High uncertainty with Standard capital allocation. Hold through the CapEx cycle.

ASML has the best moat of any publicly traded company — a physics-constrained EUV monopoly where competition is literally not feasible. The business is not in question. The valuation is. At 11% above fair value, this is the only position in the portfolio with negative margin of safety. Target trim from 9.2% to 5%, deploying proceeds into names with 25-60% discounts. The moat justifies a permanent core position — just not at a premium.

MELI is the best expression of LatAm's digital transformation — a dominant two-sided marketplace with financial services embedded at the core. Network effects compound across commerce, payments, lending, and logistics simultaneously. Currency risk and credit cycle risk are the primary uncertainties. At 22% discount with Exemplary capital allocation and 22%+ forward revenue CAGR, appropriately sized at 5.7%.

MSFT is the highest conviction position that is simultaneously the most underweight relative to conviction. 9.5/10 moat, Medium uncertainty, Exemplary capital allocation, 38% discount, strongest bear floor (+11%/yr) of any growth stock analyzed. $625B commercial RPO provides revenue visibility no other company matches. Target 6-7%. Every new capital deployment should consider MSFT first.

TSMC is the most irreplaceable business in the AI infrastructure stack — 90% of advanced AI chips manufactured here, process technology 2-3 generations ahead. Medium uncertainty reflects business stability rather than geopolitical risk (which is unmodellable). At 21% discount, appropriately held at 5.3% with the geopolitical risk serving as the natural cap on position size.

SPGI is ranked #1 on the scorecard and one of the most underweight positions relative to conviction. Low uncertainty + Wide moat + Exemplary capital allocation + 32% discount is exceptionally rare. The S&P 500 index franchise requires zero incremental capital to maintain yet generates billions in licensing fees annually — the highest quality of earnings possible. Target weight: 6-7%. Currently 4.5%.

Broadcom's combination of custom AI silicon dominance and VMware enterprise lock-in creates a uniquely defensible position in the AI infrastructure buildout. The $100B+ AI chip revenue guide for 2027 is based on existing XPU design wins at Google, Meta, and Apple. At 37% discount with Exemplary capital allocation, currently at 4.6% — slightly undersized. Key-man risk caps concentration.

MRVL is the highest-growth semiconductor name in the portfolio but carries the most explicit moat caveat — Narrow rather than Wide. The AWS Trainium and Google TPU design wins are real but the Narrow moat warrants sizing below what a Wide moat equivalent would receive. At 3.3%, appropriately sized. Would size to 4-5% if Morningstar upgraded to Wide moat.

Mastercard has 130%+ ROIC on a near-unassailable two-sided network. 10% discount is thin but near-unassailable moat quality justifies a core position even at modest discounts. Standard capital allocation is the only genuine flag. At 3.3%, appropriately sized. Would add meaningfully at 15-20% discount. Adding Visa would complete the payments duopoly.

Arista won the AI networking standard battle and is reaping the rewards with 400G/800G deployments across every major hyperscaler. The EOS software moat creates genuine switching costs that compound over time. At 38% discount with Wide moat and Exemplary capital allocation, a quality hold. NVIDIA InfiniBand competitive threat is the key watch item.

NVIDIA's CUDA moat is real and deeply entrenched — 25 years of developer investment cannot be unwound quickly. The AI training accelerator dominance generates extraordinary financial results. Very High uncertainty reflects both the lofty valuation and genuine long-term risks from custom silicon. At 2.9%, sized conservatively for Very High uncertainty. CUDA software moat is the key monitor.

Palo Alto is the most ambitious strategic bet in the cybersecurity cluster — trying to become the operating system layer for enterprise security. Platformization is working by the NGS ARR metrics but creates near-term revenue pressure. At 38% discount with Exemplary capital allocation and Wide moat, a quality hold. The ZS + PANW + CRWD cluster represents ~9% cybersecurity exposure — monitoring for rationalization.

MPWR is a genuinely excellent analog semiconductor business with a direct AI infrastructure tailwind. The Wide moat is real. The problem is the 59× P/E — even with 18-20% revenue growth, multiple compression from current levels creates a headwind. Standard capital allocation is another mild flag. At 2.5% and 16% discount, appropriately sized but not a priority for additional capital at this valuation.

ZS has the largest discount in the portfolio at 55% below fair value. The data network effect moat is real and 120%+ NRR signals genuine switching costs. However the recent Narrow moat downgrade is material. The appropriate response is to hold current 3.2% but not add until a fresh Morningstar note confirms the moat thesis. Bear floor at +12%/yr provides meaningful downside protection.

ServiceNow's 98% gross retention is Wide moat behavior from a Narrow moat business — the tension Morningstar captures in its rating. The workflow lock-in is genuine; the AI disruption risk is also genuine. At 3.2% and 33% discount, appropriately sized for a Narrow moat High uncertainty name. Don't add aggressively — but don't sell. Earns its keep at this size through RPO visibility and retention metrics.

Amphenol is the most quality-per-unit-of-risk name in the portfolio relative to its current sizing. Wide moat, Medium uncertainty (unique among growth holdings), Exemplary capital allocation, 33% discount, and 22% 20-year average ROIC. The AI data center connector thesis adds a direct secular tailwind to an already excellent industrial compounder. Currently at 1.1% — well below the 3% target. Priority for building with FTNT proceeds and ongoing capital.

Moody's is the SPGI twin — same credit ratings duopoly, same regulatory moat, similar financial profile. Medium uncertainty and Exemplary capital allocation make this a natural financial infrastructure pair with SPGI. At 25% discount and 2.5% of portfolio, MCO is slightly undersized relative to quality. SPGI is being built more aggressively given deeper discount. Together SPGI + MCO form the core financial infrastructure cluster.

CrowdStrike is a genuinely excellent security platform — Threat Graph data moat, Charlotte AI first-mover advantage, and 120%+ NRR are all real. The concern is valuation — only 15% below fair value for a Very High uncertainty name. Hold but don't add at current prices. Would become a meaningful add at 25%+ discount.

MSCI is a pristine business — Wide Moat, Exemplary capital allocation, 61% EBITDA margins, 117% FCF conversion, and mid-90s retention on its most important revenue. The index franchise is one of the most durable in financial services: once an index dominates a segment, it's almost impossible to displace. At 0.80x fair value with a 20% discount, this is a fair entry point for a new position. I initiated at $547. The BlackRock extension to 2035 removes near-term tail risk. Primary risks to watch: ETF fee compression pace, ESG segment recovery, and active asset manager budget pressure.

DDOG is the best pure-play on cloud complexity growth and the telemetry explosion from AI/agentic workloads. 97% gross retention, 120% NRR (90th percentile in software), and the AI agents tailwind make this a compounding quality business. Standard capital allocation and consumption volatility cap conviction at 3-4% rather than 5%+. Currently at 1.7% — meaningfully undersized. Build to 3% when capital is available after higher-priority gaps are closed.

Airbnb's two-sided network with 5M+ individual hosts creates a supply moat AI-powered competitors cannot bypass — the inventory doesn't exist anywhere else to index. 63% direct traffic rate is the strongest evidence of genuine brand preference. Standard capital allocation and travel cyclicality cap conviction. At 1.4% and 27% discount, appropriately sized as a diversifying holding. Not a priority for additional capital given alternatives with better risk-adjusted returns.

Duolingo is the most speculative holding in the portfolio — extraordinary revenue growth and genuine brand love, but a Narrow moat that cannot be structurally defended against a determined competitor. The 30% discount and 28% revenue CAGR partially justify ownership. But at 1.3%, it's sized appropriately for the moat quality. Standard capital allocation and Narrow moat are the twin flags that keep this a small position. Monitor closely for moat erosion.