Triatomic Q1 2025 Private
The Q1 2025 update letter from Triatomic Capital, dated April 30, 2025, outlines the firm’s view of a turbulent macro environment—characterized by the commoditization of foundational AI models, rising tariffs, and a pull‑back in public research funding—and explains how these forces create both risks and strategic opportunities. The letter highlights two core thematic bets: (1) a shift in enterprise AI from generic large models to deep‑context, data‑driven systems that rely on extensive connectors and custom knowledge bases, exemplified by a new horizontal AI‑agent product (“Cursor for work”) targeting knowledge workers; and (2) a long‑term reinvention of data‑center architecture despite a near‑term slowdown in AI‑hardware capex, with analysis of historical investment bubbles, current AI‑hardware economics, and emerging modular chip solutions (e.g., D‑Matrix). The firm reports a portfolio of 13 companies (34% engineered biology, 33% next‑gen compute, 17% automated economy, 11% engineered materials, 6% new energy), details recent diligence activity (317 opportunities reviewed, 73 meetings, 19 advanced diligences, 2 term sheets), and announces three net‑new investments—including iPronics (high‑speed data movement), a stealth lithography venture, and the “Cursor for work” startup. Portfolio updates for existing companies (Manifold Bio, Chemify, Lyten, D‑Matrix, Elegen, Teleo, MatX) are provided with milestones and upcoming financing expectations. The letter also notes team expansion in the Bay Area, recruiting needs, and includes standard confidential disclosures and forward‑looking statement cautions.
Topics
Macro Turbulence and AI Model Commoditization
Analysis of macro‑level forces—tariffs, research funding cuts, and the rapid price decline of large language model API calls—and their impact on venture investment risk and opportunity.
Enterprise AI Shift to Deep‑Context, Data‑Driven Systems
Discussion of the transition from generic foundational models to applications that integrate proprietary data, connectors, and unified data models, enabling higher‑value, verticalized AI solutions.
Horizontal AI Agent (“Cursor for Work”) as a New Enterprise Tool
Presentation of Triatomic’s first horizontal software investment—a general‑purpose AI agent designed for knowledge workers, its market sizing, and strategic relevance to the deep‑context trend.
Limitations of Early Advanced‑Context AI Products
Recognition of constraints such as data quality, connector robustness, and ecosystem lock‑ins (e.g., LinkedIn, Salesforce) that temper the immediate productivity gains of advanced‑context tools.
Reinventing Data‑Center Architecture Amid AI‑Hardware Capex Slowdown
Evaluation of the current slowdown in AI‑hardware spending, historical analogues (railroad, fiber‑optic booms), and the long‑term headroom for capital investment in next‑gen compute infrastructure.
AI‑Hardware Cost Dynamics and Model Efficiency Trade‑offs
Breakdown of how model efficiency techniques (MoE, sparsity, low‑bit numerics, CoT/ToT) affect hardware requirements, memory bandwidth, and total cost of ownership, illustrated with concrete token‑per‑second calculations.
Portfolio Overview and Recent Investment Activity
Snapshot of the fund’s composition (13 companies, sector allocations), diligence pipeline metrics, and details of recent investments (iPronics, stealth lithography venture, Cursor for work) and existing portfolio milestones.
Team Expansion and Talent Acquisition
Announcement of ongoing recruitment for an associate/principal level investment professional in the Bay Area and a call for introductions.