Fundamental raises $255 million for AI model that tackles enterprise tabular data blind spot

Fundamental Emerges with $255 Million Funding for Tabular Data AI

An AI lab called Fundamental has emerged from stealth with $255 million in total funding to address a persistent challenge in enterprise technology: how to extract insights from massive structured datasets. The San Francisco-based company, co-founded by DeepMind alumni including CEO Jeremy Fraenkel, secured the bulk of its capital through a $225 million Series A round led by Oak HC/FT, Valor Equity Partners, Battery Ventures, and Salesforce Ventures 1

1

. Hetz Ventures also participated, alongside angel investors including Perplexity CEO Aravind Srinivas, Brex co-founder Henrique Dubugras, and Datadog CEO Olivier Pomel. The funding values the company at $1.2 billion 1

1

.

Large Tabular Model Breaks from Transformer Architecture

Fundamental's flagship product, Nexus, represents a departure from contemporary AI development. Rather than a Large Language Model, the company has built what it calls a Large Tabular Model (LTM) specifically designed for structured data like spreadsheets and database tables 1

1

. The foundation model for structured data doesn't rely on the transformer architecture that defines models from OpenAI and Anthropic. Instead, Nexus is deterministic, meaning it delivers the same answer every time for a given question. This AI model for tabular data was trained on billions of real-world tabular datasets using Amazon SageMaker HyperPod 2

2

.

Why Traditional LLMs Fail at Enterprise Data Analysis

The deep learning revolution has largely bypassed the spreadsheet, leaving enterprises to forecast business outcomes using legacy machine learning algorithms that predate modern AI 2

2

. Fraenkel explained that LLMs struggle with tabular data because of how they handle tokenization. "LLMs tokenize numbers the same way they tokenize words, breaking them into smaller chunks," he told VentureBeat. "If you have a number like 2.3, the '2', the '.', and the '3' are seen as three different tokens. That essentially means you lose the understanding of the distribution of numbers" 2

2

. Additionally, transformer-based models can only process data within their context window, making them ineffective for analyzing spreadsheets with billions of rows 1

1

.

Nexus LTM Operates at the Predictive Layer

While recent integrations like Anthropic's Claude appearing in Microsoft Excel suggest LLMs are already solving tables, Fundamental operates at a fundamentally different layer. "What they are doing is essentially at the formula layer -- formulas are text, they are like code," Fraenkel explained. "We aren't trying to allow you to build a financial model in Excel. We are helping you make a forecast" 2

2

. Nexus is designed for split-second decisions where humans aren't in the loop, such as determining if a credit card transaction is fraudulent the moment a customer swipes. The model can predict customer churn, equipment failures, hospital readmissions, and store traffic patterns 3

3

.

Building Predictive Models Without Manual Feature Engineering

One of the most significant advantages of Nexus is its ability to ingest raw tables directly, eliminating the need for manual feature engineering that traditional approaches require. Unlike XGBoost or Random Forest models, where data scientists must manually define which variables the model should examine, Nexus identifies latent patterns across columns and rows automatically 2

2

. The model can automatically interpret ambiguous records, determining whether fields containing "Yellowstone" and "Yosemite" describe national parks or conference room names 3

3

. This automation drastically reduces the time required for Big Data analysis, which traditionally could take months of manual labor.

Amazon Web Services Partnership Accelerates Enterprise Adoption

Fundamental has entered into a strategic partnership with AWS that allows customers to purchase and deploy the Nexus LTM directly within their AWS environment 4

4

. "Fundamental's structured data prediction model builds on AWS' advanced AI offerings, helping enterprise customers fill a crucial gap in comprehensive tabular data analysis at scale," said Dave Brown, VP of Compute, Platforms & ML Services at AWS 4

4

. The company has already secured multiple seven-figure contracts with Fortune 100 companies for tasks including forecasting customer churn 3

3

. Fundamental plans to use the funding to scale compute resources, expand enterprise deployments, and grow its research, engineering, and go-to-market teams 4

4

.