Buying an apartment presents a small exercise in imagination. You inspect someone else’s furniture, someone else’s lighting, someone else’s regrettable taste in curtains, and try to see a life of your own. Alexandre de Vigan noticed something useful in that gap. The pictures were doing much of the selling. They were also asking buyers to perform rather a lot of unpaid mental renovation.
His response eventually became a business in making things visible. First came property. Then furniture and retail catalogs. Today, the same founder is concerned with another audience altogether: machines that need information about the physical world. A picture can persuade a shopper that a chair belongs in a room. Teaching an AI system about that room demands a different kind of representation.
De Vigan’s route between those two problems runs through a legal career, a series of changes in direction and an enterprise software business called Nfinite. In March 2026, it acquired another chapter with the launch of Physicl. The continuity lies behind the finished images: the three-dimensional objects and environments from which they can be made.
A buyer’s problem, an adviser’s impatience
Law had initially offered de Vigan a way to intervene in business. His father’s entrepreneurial difficulties helped draw him toward the profession as a child. Later came studies in law and management, including Assas, Columbia and HEC Paris. He worked on mergers and acquisitions at Darrois Villey Maillot Brochier, then also practiced at Gondran de Robert Avocats.
There is a particular vantage point available to an M&A lawyer. You see decisions with consequences, and you help people carry them out. De Vigan enjoyed the work but wanted a more direct role in the outcome. Advising companies was becoming an unsatisfactory substitute for building one. It is an understandable complaint, although starting a company is a fairly elaborate way to acquire more responsibility.
In 2014, after his own experience buying and renovating property, he founded Matchimmo, an online real estate acquisition platform. The work brought him closer to the practical business of helping someone choose a place. Imagery attracted his attention because it influenced which properties people considered in the first place. A listing could be accurate and still fail to make its possibilities apparent.
The venture that became Hubstairs grew out of this territory. De Vigan dates its launch to 2017 in his accounts of the business. Its early work included property visualization and decorating services. In 2018, it added imagery on demand for home and living retailers. The customers were changing, but the underlying question remained familiar: how could a person see an object in a setting that made sense to them?
A catalog is never finished
Property work and retail work have different rhythms. A property is a particular place. A retailer’s catalog is a moving population of products, colors, materials and campaigns. The same table might need a plain product image, a place in a furnished room and a different treatment for the next selling season. Multiply the work across a large catalog and the pleasant business of arranging a room becomes a production problem.
Nfinite’s answer was to give the product a reusable digital representation. With a three-dimensional model, a retailer could generate further views and settings without arranging another physical shoot for every variation. The model could support still images, spins and contextual scenes. The object persisted while the treatment around it changed.
That distinction made the business larger than a supply of attractive pictures. In de Vigan’s account, the real estate activity ended in 2020. The company moved toward a software platform and the Nfinite identity. He was trying to put repeatable production into the hands of retailers, with creative expertise and technology working together.
The ambition also had an ordinary technical side. In 2023, he explained that retailers could export standard image files such as JPGs and PNGs. A shopper did not need to understand the production system to encounter its output. This is one of enterprise software’s less glamorous virtues: the sophisticated machinery can remain backstage while an existing workflow keeps functioning.
De Vigan’s emphasis was on flexibility. A seasonal campaign could ask for another setting; a catalog could grow; a visual could be adapted. The appeal rested on the amount of work that would come after the first model. Making the next picture easier was the commercial proposition.
The American customer comes first
The American market gave that proposition a demanding test. De Vigan’s public account of pursuing Walmart includes years of rejection before the retailer became a customer. The objections included whether the business could operate at the required scale. That is a consequential question when the prospective buyer’s catalog is considerably larger than the founder’s capacity to charm a meeting.
Walmart, Amazon and Lowe’s became part of the customer story he described. His advice to founders considering the United States was practical: secure a customer, invest time and money, and hire people who understand American buyers. He also cautioned against treating the country as one uniform commercial culture. Arkansas and California require more thought than a shared currency might suggest.
Financing followed the business’s expansion. In February 2022, Nfinite announced a $15 million Series A led by USVP. The announcement reported tenfold growth in annual recurring revenue during 2021. In June 2022 came a $100 million Series B led by Insight Partners, with USVP participating. Nfinite said the round brought its cumulative funding to $130 million.
De Vigan described investment in catalog digitization, research and development, and commercial teams in the United States. These were the less cinematic requirements behind the growth figures. A retailer needed its products represented reliably. The platform needed to handle the work. Someone needed to sell it, and someone needed to deliver what had been sold.
The next customer cannot admire the lighting
By 2024, de Vigan was talking publicly about another use for the company’s capabilities: synthetic data for AI. The shift makes sense if you look at the ingredients rather than the finished product. A rendered room contains objects, geometry, materials, lighting and spatial relationships. For a shopper, much of that work disappears into the impression of a convincing image. For an AI developer, the underlying information can be the point.
De Vigan distinguished the result, a computer-generated image, from the methods used to produce it. Three-dimensional graphics, computer vision and machine learning could contribute in different combinations. His argument focused on control: creating particular visual conditions rather than accepting whatever happened to be present in a collection of existing pictures.
In 2025, he described his work as supplying structured environments for training and simulation. The intended users included research teams working on robotics and other systems that interact with space. That audience asks harder questions of a scene. How is an object shaped? Where is it? What changes when the viewpoint moves? A persuasive surface offers only part of the answer.
“When we founded Nfinite, the vision was clear: digitize the world in 3D.”
Alexandre de Vigan · November 2025
The aspiration was broader than retail, although retail had supplied years of practical work. In January 2026, Nfinite announced a collaboration with Getty Images to transform selected creative imagery into structured three-dimensional scenes. The proposed datasets would add information including object geometry, material properties, lighting and camera parameters. A familiar photographic library was becoming an input to another form of representation.
That collaboration also placed licensing inside the production story. Useful training material needs a workable basis for use as well as useful content. For a former business lawyer, the proximity of rights and technology is an apt detail. The commercial object is a dataset that a customer can put to work, with its provenance considered alongside its visual and spatial qualities.
A new name for an old obsession
Physicl emerged publicly at NVIDIA GTC on March 17, 2026, launched by members of the team behind Nfinite. De Vigan leads it as founder and CEO. Its declared focus is data infrastructure for physical AI: structured three-dimensional assets and environments for systems being trained to perceive and act in the world.
“Every major advance in AI has required a new data layer,” he said at the launch. His position is that progress in robotics and world models depends on improving what those systems can train on. The company’s work includes converting visual inputs into usable 3D representations and generating variations grounded in physical properties.
His introductory essay described assets carrying information such as mass, friction, collision geometry and object pose. Those details explain why a handsome digital object may still require substantial work before it belongs in a simulation. A chair in an advertisement has to look convincing. A chair in a training environment has a longer list of obligations.
Physicl’s team describes six years of working together on 3D production systems before this new company. Its first app, Studio, is designed to let teams describe a scene and generate images or videos with accompanying information such as depth and segmentation. The ambition now concerns making training material repeatable and controllable, much as the retail business concerned making visual production repeatable and controllable.
There is a pleasing personal detail in all this. Asked what he might have done instead of entrepreneurship, de Vigan chose architecture. He has also spoken about wanting to design a home by the sea. Buildings, rooms, objects and the lives arranged around them have remained close to his interests, even as the customers and vocabulary changed.
The apartment buyer wanted help imagining a room. The retailer wanted a reliable way to show its contents. The AI developer wants a world represented in enough detail to learn from. De Vigan has followed that progression into increasingly demanding work. The picture is still there. More of the business now lives behind it.