Paris has a new foundational AI player. Poolside AI, a startup that wants to create a ChatGPT-like tool that can write software code, has raised a mammoth $126m seed round, according to two sources directly involved with the fundraising.
The US-based startup has also opened a French subsidiary and decided to move its HQ to Paris, sources say. It will continue building the business in France in a big boost to the country’s AI ambitions.
French billionaire Xavier Niel and US VC Felicis led the round, which also included French business angels Rodolphe Saadé and Motier Ventures, and investors such as Bpifrance, NewWave and Frst. The round is an $100m extension to a $26m seed round announced in May, which was led by US-based Redpoint Ventures.
Poolside was founded in the US last May by Jason Warner, formerly CTO of GitHub and previously a partner at Redpoint Ventures. His cofounder, Eiso Kant, is a serial entrepreneur who has founded a number of companies including Paris-based Athenian, which provides analytics for software development.
Warner and Kant join a growing number of entrepreneurs choosing the French capital to launch and develop AI startups, attracted by the high availability and low cost of talent — including high-profile AI companies Mistral, Nabla and Dust.
The OpenAI of software development
Poolside’s team, which currently numbers less than 10, is building a foundational AI model, similar to ChatGPT but with a different objective — to eventually enable users to ask the tool, in natural language, to create code to build an application.
For example, a teacher could use the model to create an app to help their students learn, or a doctor could use it to build better software for appointment management. In other words, people with no coding experience could, in theory, become the creators of useful software.
Democratising software development in such a way could be a game-changer and creating this tool will take years. In the short-term, the company is focusing on building a model that can assist software developers in their day-to-day work, for example by optimising code or fixing bugs.
The costs of foundational AI
Poolside’s technology is based on a large language model (LLM) that will be trained on massive amounts of data; both code and natural language. Sources say that Poolside is already working with Paris-based cloud provider Scaleway to build the large GPU training clusters it will need to develop and train its own model.
Startups are currently racing to adapt to the new economy of compute, in which access to the GPUs that are necessary to train their AI models is becoming increasingly competitive.
Investing in GPU clusters and data centres doesn’t come cheap, however — partly explaining why funding rounds for foundational AI companies are so huge, even at an early stage. Just a couple of months ago, Paris-based Mistral closed a €105m seed round only four weeks after it was founded.
Poolside’s move to Paris comes a few months after French president Emmanuel Macron announced a €500m plan to encourage the development of new AI players, sending a strong signal that the country is open for business.