Fundamental Technologies, a U.S. AI startup founded in 2024, has received a strategic investment fro...

Fundamental Technologies, a U.S. AI startup founded in 2024, has received a strategic investment from MUFG Innovation Partners (MUIP), the venture arm of Japan’s largest banking group. The company is entering the Japanese market with its proprietary Large Tabular Model (LTM) called NEXUS, which is pre-trained on billions of structured data rows to generate repeatable predictions such as demand forecasting, risk analysis, and optimization. NEXUS runs inside the customer's own cloud environment (VPC) using confidential computing, ensuring no client data leaves the infrastructure.

The entry marks the first dedicated LTM play in Japan’s enterprise AI market, targeting sectors like manufacturing, retail, and telecommunications that hold decades of proprietary structured data. Fundamental claims NEXUS can reduce prediction errors by roughly 50% compared to traditionally engineered linear models, and outputs deterministic numerical scores rather than probabilistic text — a critical requirement for regulated financial and industrial applications. The company previously raised a $225 million Series A in February 2026 from Battery Ventures and Salesforce Ventures, achieving unicorn status out of stealth.

Fundamental’s move into Japan is a textbook application of the vertical-specific foundation model pattern, where a domain-specialized pre-trained model replaces the bespoke data-science workflow for structured data. This positions NEXUS as a competitor to both conventional data-science tooling and general-purpose LLMs, which the company argues process tabular data as text rather than native statistical arrays. The focus on explainability and deterministic output also speaks directly to the AI reliability debate in regulated industries, where black-box generative models have struggled to meet compliance standards.