Moats AI Can't Rebuild
Amplify@amplify
+4.4%
since Sep 28
The idea
I believe AI has made it trivial to build custom software, so the old moat of owning seats and screens is fading fast. Value is shifting to what a quick build can't replicate: the compute layer things run on, and proprietary data nobody else can access. This basket leans harder into the data side through exchanges, ratings agencies, and data monopolies, using compute and cloud names as steadier ballast, with a slice held in cash. It's built to run for years, not to trade around news.
What it owns
- Every AI application needs compute infrastructure, and the cloud and chip vendors collect revenue from all of them regardless of which software layer wins. · NVIDIAIt manufactures the dominant chips and software stack that nearly all AI training and inference workloads run on., AmazonAWS collects recurring revenue for every hour of compute that AI applications consume, regardless of which software layers win., MicrosoftAzure captures the compute spending from a large share of AI applications deployed at scale in production workloads., Taiwan SemiconductorIt manufactures nearly all leading-edge AI chips including those from competitors, collecting manufacturing revenue across the entire layer., BroadcomIt designs custom AI chips for hyperscalers seeking to reduce their spend on standard chips, winning from compute demand in either direction.29%
- Regulators and lenders require ratings and credit histories as gatekeepers to capital markets, so no software alternative can substitute for them. · S&P GlobalIt operates the ratings franchise and S&P indices that debt markets and trillions in assets require as regulatory gatekeepers and benchmarks., Moody'sIt operates the ratings half of the duopoly plus decades of credit and risk data, both locked in by regulatory requirements that AI cannot replace., Fair IsaacIt owns the credit score standard that the entire lending and mortgage system requires, a universal acceptance no generic AI build can replicate., EquifaxIt holds credit files and The Work Number employment records that lenders require and no competitor is permitted to own., TransUnionIt operates the second major credit bureau, holding files and lending data that regulators require and custom software cannot construct.24%
- Exchanges own the raw price data that every trading strategy needs, and cheaper software only increases demand for that proprietary feed. · Intercontinental ExchangeIt owns the NYSE and sells the market data that flows from it, the raw material no custom build can generate independently., CME GroupIt hosts US rates, equity index and commodity futures trading, making its prices the reference data those markets depend on., Nasdaq IncIt pairs exchange listing data with surveillance and compliance software that regulators and banks cannot replace despite cheaper trading tools., Cboe Global MarketsIt owns the options market and the VIX, and the prices from both are proprietary data that trading tools must license to operate.21%
- Decades of accumulated records in insurance, indexes, law and property remain valuable regardless of how commoditized the software layer becomes. · MSCI IncIts global and emerging-market indices are the benchmarks for trillions in institutional assets, generating fees independent of compute costs., RELX PLCIt owns legal records, scientific journals and risk databases that AI tools must license rather than replace with custom builds., Verisk AnalyticsIt operates a proprietary insurance claims database contributed by carriers in exchange for access, a data pool no outsider can assemble., CoStar GroupIt built a commercial property database through decades of field research that AI and software shortcuts cannot compress into competing time.16%
- Cash10%