Re-Catalyzing Commodities

Thesis Overview

A thematic approach to investing I’ve developed in my career has been to look at industries that exist outside of the technology paradigm where commoditization has led to compressed margins. Leading players in these markets typically persist through a combination of scale and infrastructure, rather than growth and innovation. New technologies can catalyze true competitive advantages that ride existing distribution channels to drive margin growth.

When I invested in OneCard, a $1.5B company, the idea was the incumbent Indian banking sector, the second most profitable in the world, would struggle to service growing demand for consumer credit. India has low credit bureau penetration, making underwriting difficult. While profitable, banking margins were declining due to the variable cost of scaling physical locations and administrative staff to handle the growing number of Indians joining the real economy. OneCard started with a free credit scoring and training app, OneScore, adding 7,000-15,000 new users a week. This acted as a low cost funnel into their digital credit card product, flipping the traditional consumer credit model, unlocking distribution and all without a single physical bank location required.

Today, these large-scale geographic macro themes are harder to come by. However, advancements in LLMs are priming other large but commoditized industries for change.

Opportunities

Take the hedge fund industry. Prior to Levante, I worked at Cat Rock Capital, a spin-out from Tiger Global. Alpha prior to the internet age in public markets was driven by having a meaningful edge in information that would lead investors to some golden insight, leading them to conviction. That information advantage accrued to particular hedge fund managers in different ways, and the half-life decay of information took longer because of the slower transmission of that information.

“One of the things we have learnt over the last few years is our most profitable insights have come from recognizing the deep reality of some businesses, not from being more contrarian than everyone else.” - Nick Sleep, “The Nomad Partnership Letters”

In Tiger’s early days for example, the twenty-something year old Chase, Scott and Feroz would go further than their grey haired peers, flying to China to get first-hand and not-readily-available insights into eCommerce logistics operations or telecom businesses in under-penetrated markets like Egypt. These insights led them to legendary, undiscovered investments into JD and Global Telecom Holding. At Cat Rock, the team would read through hundreds of pages of call transcripts, consumer credit card spend reports and ticker screens. The problem is that as the internet has proliferated access to this information in a market participant system where access is unconstrained, all public investors (depending on firm budgets) have the same tools and data.

That’s why we see hedge fund managers revert to marketing a focus on “fundamentals” as their edge or developing some incremental analysis of satellite imagery to make a marginal gain in commodity trading. In reality, hedge fund returns have consistently declined and as a result fees have been compressed, as a focus on fundamentals really means buying consensus cash flowing low-growth names at priced-in valuations, and data analysis focusses on low-margin trading activities that requires enormous volume to move the needle. Most of this work is also manual, time consuming and text-based.

This has led to an industry-wide problem. More available, real-time data being parsed by humans means managers need to make a conscious decision to be either quant-driven with large analyst teams like D.E. Shaw with 1,100 employees, or analysis driven with smaller teams sacrificing data-coverage like Berkshire with ~250 people. With the emergence of LLM’s that can perform human-level analysis without human-level errors, we can imagine a world where a hedge-fund employing no humans can react to datasets movements instantly, access millions-times more data than even the largest teams and work instantly across multi-modal and multi-lingual datasets 24/7.

A fully-automated hedge fund powered by LLM agents could charge 0% management fees as it requires no overhead, but could unlock order-of-magnitude higher incentive fees, re-catalyzing margin expansion.

This is a theme we believe applies to industries with similar dynamics like well discovery, formulation discovery in material sciences, film & TV production, the airline industry and more.

For example, GoNavi is tackling the global pilot shortage by integrating with audio and sensor data in cockpits to provide 24/7 instructor agents for flight schools. Pilot training is expected to be a $20B market by 2030, growing 14% CAGR, driven by new airlines in emerging markets and supply shortages of human instructors leading to 80% pilot drop-out rates. Flight schools, which employ highly specialized labor, suffer from low average gross margins of 4%. Automating cockpit command training using LLMs increases the throughput of paying students per flight school while reducing the need to increase labor linearly with student intake.