Fundamental Secures $255M in Series A for AI-driven Enterprise Predictions

Most companies talk about artificial intelligence like it is poetry. Beautiful, abstract, a little detached from the bills that need paying on Friday. Fundamental shows up talking math. Rows. Columns. The stuff executives actually lose sleep over. The spreadsheets that decide fraud, risk, inventory, patients, prices. The unsexy data that quietly moves trillions while the rest of the internet argues about prompts.

Founded in October 2024 by DeepMind alumni, Fundamental came out of stealth and did not whisper. It announced $255M in total funding, including a $225M Series A, at a $1.2B valuation. That number is loud, but the reason behind it is louder. Jeremy Fraenkel, Co-Founder and CEO, saw what most teams missed. Large language models love words. Enterprises run on tables. Those 2 worlds have been politely ignoring each other.

Enter NEXUS, Fundamental’s Large Tabular Model. Not a transformer. Not a chatbot wearing a spreadsheet costume. A purpose-built system trained on billions of tabular datasets, capable of understanding non-linear relationships across rows and columns, processing spreadsheets with billions of rows, and doing it without armies of feature engineers duct-taping models together. Built on Amazon SageMaker HyperPod, deployable directly inside existing AWS environments, speaking the native language of enterprise data instead of translating it into something trendy.

The capital came from Oak HC/FT, with Valor Equity Partners, Battery Ventures, Salesforce Ventures, and Hetz Ventures leaning in. Angels include Aravind Srinivas of Perplexity, Assaf Rappaport of Wiz, Henrique Dubugras of Brex, and Olivier Pomel of Datadog. That roster reads less like a hype circle and more like people who have scars from shipping real infrastructure at scale.

Fundamental already counts Fortune 500 companies as customers, applying NEXUS to fraud detection, hospital readmission prediction, and energy price forecasting. Three industries where being wrong is expensive, being late is fatal, and being explainable actually matters. This is not artificial intelligence for demos. This is artificial intelligence for decisions.

There is a lesson buried in this round if you slow down enough to hear it. The next wave of value is not chasing novelty. It is respecting fundamentals. Knowing where the real data lives. Building for it instead of around it. And having the restraint to say no to what is popular in favor of what is necessary.