The world of artificial intelligence has made astounding progress in understanding human language and generating realistic imagery, yet a significant blind spot persists: the vast ocean of structured, relational data that powers global commerce. Until now, enterprise data residing in spreadsheets, CRMs, and financial ledgers has been largely treated as mere file formats, forcing businesses to rely on outdated, manual processes for forecasting and prediction. That paradigm is set to shift dramatically with the emergence of Fundamental, an AI firm co-founded by DeepMind alumni, which today unveils NEXUS, a "Large Tabular Model" (LTM) designed to natively comprehend the intricate relationships within tabular datasets. This groundbreaking model promises to bypass laborious ETL processes and usher in an era of AI-driven predictive intelligence for business operations.
Bridging the Gap: From Sequential Logic to Relational Understanding
Most contemporary AI models, including the widely discussed Large Language Models (LLMs), operate on sequential logic – predicting the next word in a sentence or the next frame in a video. This sequential approach, however, is fundamentally ill-suited for the inherently non-sequential nature of enterprise data. A customer's likelihood to churn, for instance, is not a linear progression but a complex interplay of transaction history, customer service interactions, and macroeconomic trends.
"The most valuable data in the world lives in tables, and until now there has been no good foundation model built specifically to understand it," stated Jeremy Fraenkel, CEO and Co-founder of Fundamental. He elaborated in a recent interview with VentureBeat that while the AI community fixates on text, audio, and video, tables represent the most critical data modality for enterprises. Existing LLMs falter when processing tabular data due to their limitations with the scale and dimensionality of these datasets, and they struggle with numerical representations. "If you have a number like 2.3, the '2', the '.', and the '3' are seen as three different tokens," Fraenkel explained, highlighting how LLMs lose the inherent understanding of numerical distributions. Crucially, unlike language, tabular data is order-invariant; shuffling columns should not alter a prediction, a concept that traditional LLMs, sensitive to prompt order, find difficult to grasp. NEXUS, conversely, is architected to handle this inherent characteristic, treating data not as a sequence but as a web of non-linear relationships.
NEXUS: A Predictive Powerhouse for the Enterprise
Fundamental distinguishes NEXUS from superficial integrations like AI chatbots embedded in spreadsheet software, asserting that their LTM operates at a distinct "predictive layer." While tools like Claude can summarize spreadsheet content, NEXUS is engineered to predict future states – whether it's an impending equipment failure in a factory or a patient's readmission risk. This distinction is critical for applications requiring split-second, automated decision-making, such as real-time fraud detection for credit card transactions.
The company's core promise is a radical acceleration in "time-to-insight." Traditionally, developing predictive models involved months of manual feature engineering and pipeline construction, often requiring large teams of data scientists to clean and process data for each specific use case. Fundamental claims NEXUS can perform this entire process with a single line of code, leveraging its pre-training on billions of real-world tabular datasets. Trained using Amazon SageMaker HyperPod, NEXUS bypasses the need for manual feature definition that plagues algorithms like XGBoost or Random Forest, instead identifying latent patterns across rows and columns that human analysts might overlook.
Fundamental has already secured significant traction, with several seven-figure contracts with Fortune 100 companies. Their go-to-market strategy involves a partnership with Amazon Web Services (AWS), where NEXUS is available through the AWS Marketplace. This arrangement allows enterprises to leverage existing AWS credits, effectively treating predictive intelligence as a utility. For developers, the interface is Python-based, enabling direct ingestion of raw tables and labeling of target columns to initiate forecasts, with NEXUS returning regressions or classifications directly into existing data stacks. This positions NEXUS not as a conversational assistant, but as a high-speed engine for automated decision-making.
Beyond Profit: Societal Impact and Future Potential
Beyond its clear commercial applications in demand forecasting and price prediction, Fundamental emphasizes NEXUS's potential for significant societal benefit. The model is being positioned to prevent catastrophic outcomes by identifying early warning signals within structured data. Examples include predicting pipe corrosion based on sensor and maintenance data to avert crises like the Flint water crisis, or anticipating PPE shortages during pandemics by analyzing manufacturing and epidemiological data, potentially saving lives and billions in healthcare costs. Furthermore, NEXUS aims to provide advanced flood and drought predictions, as well as predict hospital readmission risks for vulnerable patient populations. "A single mother working two jobs shouldn't end up back in the ER because we failed to predict she'd need follow-up care," the company articulated, underscoring a commitment to equity through predictive intelligence.
Fundamental's ambitious vision has been backed by substantial funding: $225 million in Series A funding, led by Oak HC/FT, with significant participation from Salesforce Ventures, Valor Equity Partners, and Battery Ventures. Annie Lamont, Co-Founder and Managing Partner at Oak HC/FT, noted the "significance of Fundamental's model is hard to overstate—structured, relational data has yet to see the benefits of the deep learning revolution." The company, with a team of around 35, is aiming to redefine the operating system for business decisions, moving from bespoke models to a foundation model era for tables, akin to what ChatGPT did for text. The partnership with AWS also addresses a critical enterprise concern: data privacy. Through a complex engineering feat, Fundamental and AWS have enabled the deployment of fully encrypted models directly within a customer's environment, ensuring data remains with the client. If NEXUS lives up to its promise, it will signify a pivotal moment where AI truly masters the structured data that governs our world, shifting the focus from past analysis to future prediction.