An AI lab called Fundamental emerged from stealth on Thursday to offer a new underlying model for solving an old problem: how to extract insights from the vast amounts of structured data that companies generate. The company believes that by combining older systems of predictive AI with more modern tools, it can reshape the way large enterprises analyze data.
“LLM is great for processing unstructured data like text, audio, video, and code, but it’s not very good at processing structured data like tables,” CEO Jeremy Fraenkel told TechCrunch. “With our model Nexus, we built the best underlying model to process that type of data.”
The idea has already generated significant interest from investors. The company has raised $255 million in funding at a $1.2 billion valuation and is coming out of stealth. The bulk of that comes from a recent $225 million Series A round led by Oak HC/FT, Valor Equity Partners, Battery Ventures, and Salesforce Ventures. Hetz Ventures also participated in the Series A, with angel funding from Perplexity CEO Aravind Srinivas, Brex Co-Founder Henrique Dubugras, and Datadog CEO Olivier Pomel.
Fundamental’s Nexus, called a large-scale tabular model (LTM) rather than a large-scale language model (LLM), departs from modern AI practices in a number of important ways. This model is definitive. This means that you will receive the same answer every time you are asked a particular question. It also does not rely on the transformer architecture that defines modern AI lab models. At Fundamental, we call this a foundational model because we go through the usual steps of pre-training and fine-tuning, but the results are very different from what our clients get when they partner with OpenAI or Anthropic.
These differences are important because Fundamental pursues use cases where modern AI models often fail. Because Transformer-based AI models can only process data within the context window, they often have trouble inferring very large datasets (for example, when analyzing spreadsheets with billions of rows). However, these kinds of large structured datasets are common in large enterprises, creating huge opportunities for models that can scale.
According to Frenkel, this is a huge opportunity for fundamentals. Nexus allows the company to bring the latest technology to big data analytics, offering more power and flexibility than the algorithms currently in use.
“Now that we can use one model for all use cases, we can significantly expand the number of use cases we work on,” he told TechCrunch. “And in each of these use cases, you get better performance than you could do with an army of data scientists.”
This promise has already resulted in a number of high-profile deals, including seven-figure deals with Fortune 100 customers. The company also has a strategic partnership with AWS that allows AWS users to deploy Nexus directly from their existing instances.
