Fundamental, an AI research lab, came out of stealth on 5 February 2026 with $255m in funding: a $225m Series A led by Oak HC/FT, on top of a $30m seed round. Valor Equity Partners, Battery Ventures, Salesforce Ventures and Hetz Ventures also invested, according to the company's release published by Amazon.
Angel investors named in the release were Aravind Srinivas of Perplexity, Assaf Rappaport of Wiz, Henrique Dubugras of Brex and Olivier Pomel of Datadog. The release did not disclose a valuation, though Calcalist reported one of $1.2bn.
“NEXUS is the OS for business decisions.”Jeremy Fraenkel, CEO and Co-founder, Fundamental
What Fundamental does
Fundamental's product is NEXUS, which it describes as its first publicly available large tabular model. Tabular data is the rows and columns that sit in databases and spreadsheets. SiliconANGLE reported, citing TechCrunch, that NEXUS is not based on the transformer architecture used by most large language models.
The release said the model was trained on Amazon SageMaker HyperPod and is available directly from the company and to AWS customers through their dashboard. SiliconANGLE said it was trained on billions of tabular datasets, interprets ambiguous records without manual data preparation and generates forecasts such as equipment failure timing, flood prediction and store traffic.
The company's website describes NEXUS as pre-trained on large amounts of tabular data and then applied to a company's own data. Oak HC/FT, in its own announcement, said NEXUS uses a transformer architecture designed for numerical and relational data and needs minimal preprocessing, unlike gradient boosting methods that require manual feature engineering. It cited use cases including fraud, hospital readmissions, demand forecasting and energy prices. The company's website highlights case studies in energy demand, credit risk, customer churn and sports analytics, and says the model can be deployed in private cloud environments, including on AWS.
Customers and team
Fundamental said it has signed seven-figure contracts with Fortune 100 enterprises for uses including demand forecasting, price prediction and customer churn. Those figures are company claims. Calcalist, which named Gabriel Suissa, an Israeli entrepreneur, as a founder alongside Fraenkel, said the company has a Tel Aviv presence and Israeli research and development activity. The release said the company was founded in October 2024 by DeepMind alumni, with Jeremy Fraenkel as chief executive and co-founder.
Oak HC/FT said Fraenkel has a graduate degree in machine learning from UC Berkeley and previously worked in finance at JPMorgan and Bridgewater. The sources do not state a headquarters city or a headcount. The release carried a San Francisco dateline.
What the money is for
AWS customers can buy and deploy NEXUS through their AWS dashboards, and Oak HC/FT said Fundamental is a first-party seller of record there. Oak said it invested because it believes large tabular models have matured enough to unlock value from enterprise data that language models have not reached, and cited an estimate that more than 80% of enterprise data is stored in tables.
The funding will go to scaling compute, expanding enterprise deployments, and growing research, engineering and go-to-market teams, the release said.
Fraenkel said NEXUS is the operating system for business decisions. Annie Lamont, co-founder and managing partner of Oak HC/FT, said the significance of Fundamental's model is hard to overstate. Dave Brown, vice president of compute, platforms and machine learning services at AWS, said Fundamental's structured data prediction model builds on AWS's advanced AI offerings.
Company website: fundamental.tech