Triatomic Q4 2025 - Private
The letter, dated January 30 2026, provides Triatomic Capital’s Q4 2025 update for its private funds. It begins with macro‑level observations: hyperscaler AI‑infrastructure capex hit a record $536 B in 2025 (up 70% YoY), AI applications are moving beyond chatbots into enterprise domains, and SaaS valuation multiples have compressed while hardware/infrastructure moats are re‑pricing positively. The authors argue that AI compute is shifting from a GPU‑centric model to a heterogeneous, rack‑scale architecture, citing changing hardware mixes at firms like Anthropic and recent strategic moves by Meta, Nvidia, OpenAI, and Anthropic. They highlight the rise of “system‑level” performance and the opportunity for startups that can abstract and orchestrate across diverse chips. The second macro theme covers “agentic AI.” While enterprise pilots are proliferating, revenue remains concentrated in a few job functions (coding, IT) and verticals (healthcare, legal). Studies show most generative‑AI pilots fail to deliver economic gains, suggesting a gap between narrow task automation and end‑to‑end workflow integration. The letter posits that longer‑horizon agents with richer memory and tighter data integration will unlock the next wave of value. Portfolio‑level details follow: the fund holds 20 companies, 11 core investments (> $10 M lifetime commitment). Allocation is 37% next‑gen compute, 28% engineered biology, 23% automated economy, 9% engineered materials, 4% new energy. Recent activity includes 205 opportunities screened, 46 meetings, 19 advanced diligences, and one new core investment expected to close in Q1 2026. Specific company updates are tabulated, noting term sheets, fundraising rounds, commercial milestones, and potential exits (e.g., MatX Series B, Manifold Bio Roche deal). Operational updates: addition of Rosa Jahankhah to the investment team, launch of marketing for Triatomic Capital Private II LP with upcoming closes, and expansion of the advisory roster. The letter concludes with standard legal disclosures and forward‑looking statement cautions. Overall, the document blends market analysis, strategic thesis on compute heterogeneity and agentic AI, concrete portfolio performance metrics, and fund‑raising progress, aimed at informing existing investors of risks, opportunities, and near‑term expectations.
Topics
AI Infrastructure Arms Race: Record Hyperscaler Capex and Shifting Valuations
Analysis of 2025 AI‑infrastructure spending ($536 B, +70% YoY) and its impact on SaaS and hardware valuation multiples.
From GPU‑Dominant to Heterogeneous Rack‑Scale Compute
Discussion of the transition from GPU‑centric AI clusters to diversified, system‑level architectures, illustrated with Anthropic’s evolving hardware mix and corporate strategy shifts at Meta, Nvidia, OpenAI, and Anthropic.
Agentic AI in the Enterprise: Early Adoption, Revenue Concentration, and Productivity Gaps
Evaluation of enterprise AI adoption, revenue distribution across job functions and verticals, and evidence that most pilots fail to deliver economic gains, leading to a hypothesis about next‑generation agents with deeper integration.
Portfolio Composition and Core Investment Metrics
Breakdown of the fund’s 20‑company portfolio, allocation percentages across thematic pillars, and core‑investment criteria (> $10 M lifetime commitment).
Company‑Level Updates and Deal Activity Snapshot
Tabular updates on key portfolio companies (e.g., MatX Series B, Manifold Bio Roche deal, Edison Scientific’s Kosmos platform) and recent fundraising, M&A, and term‑sheet activity.
Fund‑II Marketing Launch and Capital Call Timeline
Announcement of Triatomic Capital Private II LP marketing, anticipated first and subsequent closes, and upcoming capital call schedule.
Team Expansion and Talent Acquisition
Introduction of new associate Rosa Jahankhah, her academic background, and expected contributions to sourcing and technical diligence.