Second-Order AI Beneficiaries
Jesse Venticinque@jesse_
+3.2%
since Oct 3
The idea
I think the market has already priced the obvious AI winners, so this take hunts for the next layer: companies where AI will materially change the business within five years, but the stock isn't yet being traded as an AI story. It owns 13 names split across three pathways, labor-heavy firms whose costs AI can compress, quiet suppliers whose products AI buildouts consume more of, and owners of data and rights AI must license.
What it owns
- Suppliers of components and infrastructure that AI buildouts consume in larger quantities, still priced as ordinary cyclicals without AI recognition. · AtkoreIt supplies electrical conduit and cable management that data centers require in miles per facility, with volumes and pricing recoverable from post-COVID lows as AI construction scales., Mueller IndustriesIt makes copper tube and fittings for liquid cooling and electrical systems in data centers, where unit volume growth in piping and industrial segments matters more than copper prices., Valmont IndustriesIt manufactures transmission towers and utility structures for grid connections that data centers require years before AI demand becomes obvious, tracked through utility segment backlog and revenue., LittelfuseIt makes circuit protection and power semiconductors consumed more densely in higher-power AI server racks, with industrial and electronics segment revenue growth outpacing the broader cycle., Regal RexnordIt supplies motors and drives inside cooling and air-handling equipment that AI facilities purchase in larger quantities, with data-center-adjacent orders in Industrial Powertrain and Automation segments as the test.31%
- Labor-intensive businesses where AI automation directly compresses operating expenses and improves efficiency metrics. · Marsh (formerly Marsh & McLennan)It automates policy comparison, benefits analysis, and claims advocacy through announced AI programs, with revenue per employee and Risk and Insurance Services margins as tests of whether headcount truly stays flat., Progressive CorporationAI reads damage photos and settles simple claims with fewer adjusters, directly affecting its largest controllable cost, visible in the combined ratio and expense ratios versus peers., CitigroupIt employs tens of thousands in operations, compliance, and back-office roles that AI can reduce, with the efficiency ratio target already disclosed and falsifiable against actual execution., McKesson CorporationIt distributes pharmaceuticals through a high-volume, low-margin model where AI-driven efficiency gains in routing and labor productivity flow directly to profitability.30%
- Companies controlling proprietary data and content that AI models must license or acquire, while market attention remains elsewhere. · Verisk AnalyticsIt owns decades of proprietary insurance loss data that AI underwriting models must license, creating subscription and pricing tailwinds as AI adoption spreads through the industry., CoStar GroupIt controls decades of commercial property records that AI real estate tools must license rather than scrape, with information and analytics revenue acceleration as AI adoption spreads., Intercontinental ExchangeIt owns exchange and fixed-income data that trading and analytics AI consumes more of, with data services recurring revenue acceleration above historical trend as the confirmation., John Wiley & SonsIt owns deep scientific publishing archives that AI model builders need as vetted training data, with licensing revenue already signed and reportable as a distinct, growing line item.29%
- Cash10%