Fundamental, an AI lab that has recently come out of stealth mode, has introduced a pioneering foundation model aimed at addressing a longstanding challenge: extracting valuable insights from the vast amounts of structured data generated by enterprises. By integrating traditional predictive AI systems with cutting-edge tools, the company aims to redefine data analytics for large organizations.
“While large language models (LLMs) excel at processing unstructured data, such as text, audio, video, and code, they struggle with structured data like tables,” stated CEO Jeremy Fraenkel in an interview with TechCrunch. “Our model, Nexus, represents the pinnacle of foundation models tailored to manage such data effectively.”
The innovation has attracted considerable interest from investors. Emerging from stealth mode, the company has secured $255 million in funding, achieving a valuation of $1.2 billion. The majority of this funding was raised in a recent $225 million Series A round led by Oak HC/FT, Valor Equity Partners, Battery Ventures, and Salesforce Ventures, with additional participation from Hetz Ventures and angel investments from notable figures including Perplexity CEO Aravind Srinivas, Brex co-founder Henrique Dubugras, and Datadog CEO Olivier Pomel.
Referred to as a large tabular model (LTM) instead of a large language model (LLM), Fundamental’s Nexus departs from conventional AI paradigms in several critical aspects. The model is deterministic, ensuring consistency in responses to identical queries, and eschews the transformer architecture that characterizes most modern AI models. Fundamental defines it as a foundation model because it undergoes the standard processes of pre-training and fine-tuning, resulting in a product that is significantly different from what clients would obtain from engaging with OpenAI or Anthropic.
These distinctions are crucial as Fundamental targets a use case where existing AI models frequently encounter challenges. Transformer-based AI systems are typically limited by their context windows, making them ill-equipped to reason over extremely large datasets—such as analyzing spreadsheets containing billions of rows. Such expansive structured datasets are commonplace in large enterprises, presenting a substantial opportunity for models capable of managing this scale.
According to Fraenkel, this represents a tremendous opportunity for Fundamental. With Nexus, the company can apply contemporary methodologies to big data analysis, offering a solution that is more potent and adaptable than the current algorithms in use.
“Now, you can utilize a single model for a wide range of applications, significantly broadening the scope of challenges you can address,” he asserted to TechCrunch. “For each application, you achieve superior performance compared to what could be accomplished by deploying a large team of data scientists.”
This potential has already attracted several high-profile contracts, including seven-figure agreements with Fortune 100 companies. Additionally, the firm has forged a strategic partnership with AWS, facilitating the direct deployment of Nexus from existing AWS instances for its users.
About the Author
