Will the market of foundation models be more fragmented than in 2024 by EOY 2026?
➕
Plus
3
Ṁ80
2026
54%
chance

On the Cognitive Revolution podcast (28 June 2024), Martin Casado from a16z predicts that the market of foundation models will be more fragmented than people think, owing to decreasing marginal returns stemming from model distillation methods. According to Casado, the benchmark market structure in foundational models is an oligopoly, the running analogy being the software industry.

To operationalise the bet, I follow Jai Vipra and Anton Korinek's empirical analysis in "Market concentration implications of foundation models" (2023, Brookings WP) to give some numbers to market shares of leading commercial LLM chatbots using traffic data from similarweb.com. I have included

Data refer to June 2024 (monthly visits).

Question resolves YES if by EOY 2026, the market share of the first and second "foundation models" will differ by less than 60 percentage points. Market share is parametrised by the share of monthly visits averaged during December 2026, using similarweb.com or comparable services, where the total number of foundation models allowed in the pool cannot exceed ten.

If by EOY 2026 no consensus emerges as to what a "foundation model" is, e.g. due to advancements in AI technologies, the question resolves NA. The working definition used by Vipra and Korinek (2023) is "large artificial intelligence (AI) models that can be adapted for use in a wide range of downstream applications".

60 pp as a target: keeping the software industry as a running benchmark, the difference between Windows and macOS in the global market share held by operating systems for desktop PCs has stabilised since 2020 below 60 pp (source: https://www.statista.com/statistics/218089/global-market-share-of-windows-7/).

Get
Ṁ1,000
and
S3.00
© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules