Triatomic Q3 2024 Private
The Q3 2024 investor letter from Triatomic Capital, dated October 31 2024, outlines the firm’s macro‑level investment theses and recent activity. The authors emphasize the accelerating shift to next‑generation compute (GPU‑centric datacenters) and the growing importance of engineered biology, while expressing caution about over‑hyped AI‑enabled enterprise applications. A central narrative is the “data wall” – the projected scarcity of high‑quality human‑generated text for training ever larger LLMs – and three mitigation pathways: data preprocessing/quality, synthetic/recursive techniques (including Chain‑of‑Thought), and vertical/domain‑specific data integration. The letter illustrates how physical‑world data collection fuels human augmentation, citing autonomous vehicle ADAS evolution, surgical robotics, and Teleo’s semi‑autonomous heavy‑equipment platform. It then details recent fund activity, highlighting new and existing portfolio companies (e.g., Teleo, Manifold Bio, Chemify, BetterOmics, Lyten, dMatrix, Elegen) and their milestones. The piece concludes with team‑building updates, fund‑specific disclosures, and a reminder of the confidential nature of the communication. Throughout, the tone is formal, analytical, and cautiously optimistic, balancing enthusiasm for emerging opportunities with skepticism about unsustainable hype.
Topics
Next‑Gen Compute & GPU‑Dominant Datacenters as a Core Investment Thesis
Discussion of the industry‑wide shift toward GPU‑centric data center architectures, including implications for power delivery, location, and software stacks.
Engineered Biology – Exciting Yet Capital‑Constrained Frontier
Overview of the engineered biology sector, noting high scientific promise but limited funding and slower deal flow.
Skepticism Toward AI‑Enabled Enterprise Applications
Argument that many AI‑driven enterprise startups lack sustainable data moats and that the firm is applying a high bar to investments in this space.
Physical‑World Data Collection as a Driver of Human Augmentation
Illustrates how gathering real‑world sensor and interaction data (e.g., ADAS, surgical robots, Teleo) enables hybrid automation and improves safety, productivity, and fleet utilization.
Autonomous Vehicle Timeline & ADAS Trick‑Down Effect
Chronology of missed Tesla autonomous milestones and the resulting acceleration of driver‑assist technologies, with price‑trend tables showing democratization of ADAS features.
Surgical Robotics Data Flywheel
Analysis of how Intuitive Surgical’s Da Vinci platform generates massive procedural data, creating opportunities for precision AI models in medicine.
Teleo – OEM‑Agnostic Semi‑Autonomous Heavy‑Equipment Platform
Case study of Teleo’s hardware/software solution that digitizes operator inputs, delivering productivity gains, safety improvements, and new labor‑access models.
The ‘Data Wall’ – Projected Scarcity of Human‑Generated Text for LLM Training
Quantitative exposition of historic LLM training dataset growth, extrapolation of future demand versus global text supply, and the risk of diminishing returns.