Fundamental, an AI start-up based in San Francisco, California, has exited stealth mode with the public launch of its large tabular model (LTM), NEXUS, while raising $255m in funding.

Founded in October 2024, the company secured a $30m Seed round and a $225m Series A round led by Oak HC/FT, with additional investment from Valor Equity Partners, Salesforce Ventures, Battery Ventures, and Hetz Ventures.

Notable angel investors include Aravind Srinivas of Perplexity, Assaf Rappaport of Wiz, Henrique Dubugras of Brex, and Olivier Pomel of Datadog.

The capital is allocated for scaling computing capabilities, enhancing enterprise deployments, and expanding teams across research, engineering, and market development.

Alongside its public launch announcement, Fundamental has entered into a partnership with Amazon Web Services (AWS). This agreement enables AWS customers to purchase and implement NEXUS directly through their AWS environment.

This collaboration aims to facilitate the adoption of NEXUS among enterprises by integrating with existing AWS services.

“Fundamental’s structured data prediction model builds on AWS’s advanced AI offerings, helping enterprise customers fill a crucial gap in comprehensive tabular data analysis at scale,” said

AWS compute, platforms and machine learning services vice president Dave Brown said: “By partnering with Fundamental, we are making it seamless for customers to transform tabular data – the backbone of enterprise decision-making – into a powerful predictive asset. This collaboration exemplifies our commitment to bringing transformative AI solutions to market with the enterprise-grade security and scalability our customers demand.”

Fundamental has secured contracts with various Fortune 100 companies that plan to use NEXUS for predictive analytics in demand forecasting, price prediction, and customer churn analysis.

NEXUS is designed by alumni of DeepMind to focus on tabular data rather than unstructured data like text or images. It addresses limitations of algorithms that are often not suited for complex relationships in enterprise-scale tabular datasets.

The model processes raw data tables from various industries without extensive feature engineering or manual training processes.

Fundamental CEO and co-founder Jeremy Fraenkel said: “Over the past three years, we’ve watched AI transform how we interact with unstructured data like text, audio, images and video. ChatGPT can write poetry. Sora can produce movies.

‘Claude can code. But the most valuable data in the world doesn’t live in documents, codebases or pixels. It lives in tables – spread across databases and operational systems that power the world’s largest companies and governments.”

The introduction of NEXUS provides enterprises with a foundation model that simplifies predictive analytics tasks. Its integration is said to require minimal effort within existing IT infrastructures due to its architecture that autonomously learns underlying patterns in raw data.

NEXUS is designed to make AI-driven decisions accessible across sectors without requiring custom model developments traditionally needed by machine learning applications. This approach aims to derive actionable insights efficiently through tabular processing capabilities.