Dutch Researchers Build AI Tool That Cuts Through One Trillion Home Renovation Options to Find Your Best Path to Energy Efficiency
Eindhoven, Tuesday, 9 June 2026.
A TU/e-developed AI tool, launched on June 4, 2026, uses machine learning trained on 350,000 scenarios to navigate over one trillion renovation combinations — delivering every Dutch homeowner a personalized, affordable route to a greener home.
A Problem Eight Years in the Making
The story of the Renovation Explorer begins not in a laboratory, but around a negotiating table. In 2018, during the Dutch national climate agreement negotiations, a fundamental challenge came into sharp focus: how could the Netherlands cost-effectively scale up home renovations to meet its climate commitments, when the sheer complexity of renovation choices was paralyzing the very homeowners the policy was designed to help [1][2]. That question set in motion a research effort that would take eight years, multiple rounds of funding, and a team of specialists from Eindhoven University of Technology (TU/e), based in Eindhoven, the Netherlands, to resolve [1][2][3].
From a €30,000 Prototype to a €3 Million Machine Learning Platform
The funding journey behind the Renovation Explorer charts a methodical escalation of ambition and resources. In 2019, the project received €30,000 to construct a first working prototype over three months [1][2]. A year later, in 2020, a second, more sophisticated prototype was built over eight months using €100,000 in funding [1]. By the summer of 2021, a formal project tender was launched with partners TNO and Smart Twin [1][2]. Then, in April 2022, the project officially commenced its current phase, backed by €3 million from the Netherlands Enterprise Agency — known in Dutch as Rijksdienst voor Ondernemend Nederland, or RVO [1][2][3]. That investment, representing a funding increase of 2900 percent over the 2020 prototype budget, enabled the assembly of a full research team of postdoctoral researchers at TU/e [1][3].
How the Tool Actually Works
At its technical core, the Renovation Explorer combines physics-based energy balance simulations with a machine learning model trained on 350,000 scenarios, enabling it to dynamically evaluate over one trillion renovation packages — and to do so on a standard laptop, without requiring specialist hardware [1][2][5]. This is a meaningful departure from conventional tools. Rather than processing renovation options generically, the Renovation Explorer incorporates the precise characteristics of a specific home: its heating systems, construction type, and critically, the behavioral patterns of its occupants [1][2][5]. As Havinga explained to TU/e’s own publication Cursor on June 5, 2026, the tool calculates outcomes based on a home’s specific characteristics but also accounts for residents’ behavior, “such as which rooms they heat or whether they leave their bedroom windows open at night” [5].
Launch, Open-Source Commitment, and the Road to EU Adoption
On Thursday, June 4, 2026, the Renovation Explorer was formally unveiled at the Renovation Explorer Congress on the TU/e campus in Eindhoven, held between 10:00 and 15:00 [1][2][3]. The event included a live demonstration and marked the official release of the open-source version of the tool [1][2]. Havinga has been explicit about why open-source distribution is non-negotiable: “The scripts need to be available for other developers so that they can create their own custom implementations based on the Renovation Explorer. This tool just must be open source and the development of it must not be commercialized” [2]. In April 2026, the Dutch government announced continued funding to scale the machine learning tool to neighborhood-level models [1][2].