House of AI Startup Challenge 2026: Six Months of Free Membership at Groupama Immobilier's 14,000 sqm AI Campus in Paris La Défense, Closing 28 September 2026

House of AI Startup Challenge 2026: Six Months of Free Membership at Groupama Immobilier’s 14,000 sqm AI Campus in Paris La Défense, Closing 28 September 2026

Groupama Immobilier, the real estate arm of the French insurance group Groupama, has opened a startup challenge to recruit the founding cohort of the House of AI — a 14,000 square metre campus it is creating inside La Défense, Europe’s largest business district. Applications opened on 20 June 2026 and close on 28 September 2026, with winners announced on 30 October 2026.

The proposition is unusual enough to be worth stating precisely, because it is not a grant. There is no cash award attached to this challenge. What the top three winners receive is six months of complimentary membership at the House of AI, with full access to the campus, the resident community, and the corporate and investor network of La Défense. Finalists who do not win are granted access to the first batch of residents on the same terms. For an applied AI company selling into large enterprises, that is a distribution asset rather than a funding round — and whether it is worth your application time depends entirely on whether your go-to-market runs through the kind of company that has a tower in La Défense.

The organiser’s own framing on the challenge page is blunt about the thesis: “La Défense is Europe’s largest business district and runs on corporate giants, but the AI startups reshaping their industries are building elsewhere.” The House of AI is described explicitly as “not an incubator” but as “an ecosystem built where the decisions are made.”

Key Details

What the House of AI Actually Is

The campus is a conversion of existing office stock rather than a new build. French property and business press reporting from the MIPIM real estate fair identified the site as Le Diamant, a 14,000 sqm office building completed in 1992, on the edge of Boulevard Patrick Devedjian in Puteaux, near the Campus Cyber. Groupama’s own teams were reported to be vacating the building in mid-2026 to relocate to Nanterre, with the House of AI occupying the space for a period of years before the building’s expected longer-term conversion to housing.

That backstory matters for two practical reasons.

First, it explains the economics. Reporting on the announcement described rents set “well below market” — the building is in a transitional phase, so Groupama Immobilier can price occupancy as an ecosystem play rather than a yield play. This is temporary urbanism applied to an AI cluster. The complimentary six-month membership for winners sits inside that same logic.

Second, it tells you the intended tenant mix. The stated ambition is to gather startups, investment funds, corporate AI teams, training organisations, and public sector actors in one building. The reporting was specific that the target is not the large AI research labs but companies working on operational AI — AI applied to running a business — rather than research or academic AI. That is consistent with the challenge criteria, which repeatedly emphasise deployment readiness over novelty.

If you are a research-stage team whose strongest asset is a model or a paper, this is a poor fit. If you are a team whose strongest asset is three signed enterprise pilots, you are exactly who they described.

Who Should Apply

The official call is for “startups and scale-ups building applied AI solutions for real business challenges” across seven named domains: real estate, insurance, finance, industry, mobility, HR, and enterprise automation. The stage band runs from pre-seed to Series B.

Three qualifiers in the call text do real filtering work, and you should read them as requirements rather than aspirations:

• Your solution should be grounded in concrete use cases. Not a horizontal platform pitch with a list of possible applications, but a specific job in a specific function.
• It should demonstrate measurable results in professional environments. Note “professional environments” — results from a benchmark or a consumer beta will not carry the same weight as results from inside a real company’s workflow.
• It should be ready, or close to ready, for enterprise deployment. Procurement-readiness, security posture, and integration work count here.

The page also names an explicit bonus: technology that “integrates with existing corporate workflows and data infrastructure.” For an insurance or real estate buyer, integration is usually the entire adoption cost. A startup that can describe, concretely, how it plugs into an existing data estate is answering the question the corporate members of this campus will actually ask.

The official challenge page does not publish a company age cap or an employee headcount limit. Third-party listings of this challenge have circulated figures for both; those figures are not stated on the official page, so treat them as unconfirmed and check the challenge rules document linked from the application flow before assuming you are excluded on those grounds. The stage band — pre-seed to Series B — is the criterion the organiser states directly.

The Four Selection Criteria, and What Each One Is Testing

Groupama Immobilier publishes four criteria. They are worth taking literally, because they read as a scoring rubric rather than marketing copy.

Differentiated AI. The stated test is “a clear and defensible edge over existing approaches,” followed by the phrase “Real use case, real value, no AI for AI’s sake.” That last clause is a warning. A meaningful share of applications to challenges like this describe a workflow tool with a model bolted on. If the AI component could be removed and the product would still broadly work, expect a low score here. Be explicit about what the model does that a rules engine or a human process could not.

Ecosystem fit. The test is “true appetite for collaboration with corporates, investors and fellow residents.” This is a residency, and the organiser is selecting for people who will actually show up and participate. A founder who wants a desk and no meetings is not the target. If you have a track record of corporate pilots, joint development agreements, or co-selling relationships, that is the evidence to cite.

Growth and maturity. The test is “scalable model, ambitious market. MVP or live solution with early traction or field validation.” An MVP is explicitly acceptable — you do not need revenue — but it must exist and it must have been exposed to a real user. Field validation is the floor.

Team quality. The test is founders and key profiles who “combine technical and sector expertise and a proven execution track record.” Note the word combine. A team of strong engineers with no insurance, real estate, or industrial background is weaker against this rubric than a team that pairs both. If your sector expertise sits with an advisor or an early customer rather than a founder, say so and make the relationship concrete.

Timeline and How the Process Runs

The published timeline has three fixed points:

• 20 June 2026 — applications open
• 28 September 2026 — applications close
• 30 October 2026 — winners announced

Roughly a month separates the close of applications from the announcement, which suggests a compressed review and shortlisting process rather than a long multi-round assessment. The rewards text distinguishes between “the top-3 winning startups” and “finalists,” so a shortlist stage exists between submission and the final decision. The official page does not publish the number of finalists, the format of any pitch or interview stage, or whether a jury sits in person. Plan for the possibility of a short-notice pitch in October and do not treat submission as the end of the work.

The application is hosted on Viva Technology’s Startup Challenges and Awards platform, which runs many corporate-sponsored challenges on a shared submission system. Two consequences follow. First, applying is free under the platform’s standard terms. Second, you will be filling in a structured form rather than uploading a free-form deck, so the quality of your short-answer writing matters more than the quality of your design.

One caution specific to this platform: because the challenge pages are templated, some boilerplate from other sponsors’ challenges can appear in secondary sections of the page. If you encounter text describing business sectors or data controllers that have nothing to do with Groupama Immobilier or the House of AI, that is template residue, not a requirement of this challenge. The authoritative content is the challenge description, the four selection criteria, the rewards, and the timeline.

Preparing a Competitive Application

Given the criteria, the highest-leverage preparation is assembling evidence rather than polishing narrative.

Pick one use case and lead with it. The rubric rewards concreteness. Choose the single deployment you can describe in the most operational detail — which function, which system it touches, what the before-and-after numbers were — and make that the spine of the application. Breadth can appear later as a market-size argument.

Quantify results from professional environments. Percentage reduction in handling time, error rates, cost per case, cycle time. If your pilot was small, say it was small and give the real number rather than an extrapolation. Reviewers in insurance and real estate assess pilots for a living and will discount inflated figures faster than they will discount modest ones.

Name your integration surface. Which systems of record do you connect to? What does deployment inside a large French corporate actually involve — data residency, SSO, security review, procurement path? This directly answers the stated bonus criterion and is the part most applicants skip.

Make the collaboration case, not just the product case. Ecosystem fit is a full quarter of the rubric. Be specific about what you would do with daily access to the corporates next door: which functions you want to meet, what a pilot would look like, what you would bring to other residents.

Be honest about stage. Pre-seed is explicitly in scope. Overstating traction to sound Series B is a poor trade against a rubric that already accepts an MVP with field validation.

Common Mistakes

• Applying as a research or model company. The programme’s stated target is operational, applied AI. A frontier-model or academic pitch is misaligned with both the criteria and the tenant mix.
• Treating it as a funding opportunity. There is no cash prize on offer. If your application is written as a request for capital, it is answering a question nobody asked.
• Ignoring geography. The entire value of the reward is physical proximity to La Défense. If nobody on your team can plausibly be in Puteaux with any regularity, the membership is worth very little to you, and that will likely show in your ecosystem-fit answer.
• Generic sector claims. “We serve financial services” is weaker than “we serve claims handlers at mid-size French P&C insurers.” The seven named domains are an invitation to be specific within one, not to claim all of them.
• Leaving the pitch stage unplanned. With winners named on 30 October and applications closing 28 September, any shortlist contact will come quickly. Have a short deck and a demo environment ready before you submit.

Frequently Asked Questions

Is there any money attached? No. The published reward is six months of complimentary membership at the House of AI for the top three winners, plus access to the first resident batch for finalists. No cash prize is stated on the official page.

Does Groupama Immobilier take equity? No equity stake is mentioned in the challenge description or the rewards section.

What happens after the six free months? The official page does not state the terms of continued membership. Reporting on the House of AI announcement described rents for the building as set well below market, but no published rate applies specifically to challenge winners. Ask during the process.

Do I need to be a French company? The challenge page does not state a nationality or incorporation requirement. The practical constraint is that the reward is physical access to a campus in Puteaux, so international applicants should think about whether they can use it.

Is applying free? Yes, under the standard terms of the VivaTech Startup Challenges platform.

How many startups will be selected? Three winners are named explicitly. The number of finalists granted access to the first resident batch is not published.

When does the campus open? Reporting from the announcement pointed to an opening in autumn 2026, aligned with the 30 October winners announcement. The official challenge page does not publish an opening date.

Official Links and Next Steps

The application, the full selection criteria, the timeline, and the challenge rules are all on the official challenge page:

• Challenge and application page: https://challenges-awards.vivatech.com/challenges/groupama-immobilier-2026
• Organiser: Groupama Immobilier — https://www.groupama-immobilier.com/

If you are in scope — applied AI, pre-seed to Series B, one of the seven named domains, with something real running in a professional environment — the application cost here is low and the deadline of 28 September 2026 leaves room to gather evidence properly. Spend that time on the pilot numbers and the integration story rather than on the deck. The rubric is asking whether corporates in La Défense could deploy you next quarter, and every one of the four criteria is a different way of asking the same question.