The AI Holding Company That Builds Instead of Buys

01 /The Concept

The headlines around today's startup read well: "World's first AI holding company Infinity Constellation set to disrupt the $2 trillion professional services market"

The coverage follows a $24M raise – on top of the $12M the company raised almost exactly a year earlier.

As usual, "world's first" is a stretch, so let's focus on what the actual model is.

Infinity Constellation's thesis: in professional services, the future is "build," not "buy."

Private equity has been doing the opposite – buying professional services firms and rolling them up into portfolios. The thesis there is also AI-driven: deploy AI into acquired firms, boost efficiency and margins, capture the multiple expansion.

Infinity Constellation agrees that AI will disrupt legacy SaaS and old-school outsourcing in professional services. But it argues the PE playbook requires too much capital and takes too long to pay off. The better path: build new companies from scratch, AI-native from day one, led by experienced founders.

Those founders receive 25% of each new company – companies Infinity believes can reach $1B in value. The preference is for "serial" founders who have already built and scaled companies, ideally successfully.

Infinity stays out of day-to-day operations, but provides founders with three things:

- Access to a core software platform for data processing and AI model integration. - A shared back-office covering legal, finance, and go-to-market functions. - "World-class" expertise from the parent company's specialists and its extended network.

The expertise claim holds water: Infinity Constellation was built by Invisible Technologies, a company that has trained AI models for a large share of the world's leading AI labs and has raised $107.9M in the process. The shared infrastructure is battle-tested, not theoretical.

The model rests on three pillars – and what makes them work together is what sets this apart from a typical corporate accelerator:

- Build new companies from zero. - A shared "collective brain" across all portfolio companies – common data and proven know-how. - Network effects: each new company makes the next one faster, cheaper, and easier to launch, because the shared brain is already trained.

Infinity claims new portfolio companies generate their first revenue within one month of founding and hit a $1M annualized revenue run rate within two months.

So far, $10M has been invested across 8 companies (listed on the website). The goal is 20 by year-end. Combined, those companies have reportedly already generated $7M in revenue.

02 /Why It Matters

The idea for the holding company emerged because Infinity's founder kept encountering promising AI product ideas that didn't fit Invisible Technologies' core strategy. Rather than keeping those ideas on a shelf, he built a structure where experienced operators could take them on.

Their job: operationalize the idea, recruit the right people, and leverage the shared platform. The critical distinction from an internal "new direction" or corporate innovation lab is that this avoids defocus, conflict of interest, and operational chaos.

Two structural shifts make this model viable now, in ways it wouldn't have been before.

First: each company used to require substantial headcount to be taken seriously. Now companies can hire AI – making each portfolio company far more manageable and capital-efficient. A holding company leader can align on a vision with a founder and trust that AI will execute the plan without the coordination overhead that used to sink these structures.

Second: shared infrastructure used to mean shared back-office (accounting, HR) – helpful but not transformative. AI-native companies, however, are built on strikingly similar architectural principles. What varies is domain-specific: which agents to train, what sequence to connect them in. The platform layer – creating, training, and monitoring AI agents – is essentially the same across companies. The shared brain becomes a genuine strategic asset, not just a cost-sharing arrangement.

That's why portfolio companies hit $1M ARR in two months. Without real synergy at the platform level, that wouldn't be possible.

The AI holding company concept is circulating in different forms.

Feltsense ([related review](/review/zarabatyvat-100-millionov-dollarov-v-god)) raised $5.1M in February for a holding company designed to launch "thousands" of software products – with AI founders building each one, with the ability to bring in humans for tasks AI still handles poorly.

Rocketable ([related review](/review/v-obshhem-sluchae-jeto-poka-fantastika-a-v-chastnom-vozmozhnost-na-milliard)) raised $6.5M last summer for a holding company that acquires profitable, working products from their creators – then lets AI continue developing them.

Modus ([related review](/review/svoj-ne-za-dengi-a-za-dolju)) announced $85M in April to build a holding company that takes equity stakes in professional services firms and provides them with AI platforms and capital to grow independently.

Multiplier ([related review](/review/unikalnyj-moment-kogda-dlja-masshtabirovanija)) raised $27.5M last summer to exchange equity in professional services firms for its AI platforms – so the original owners earn more even with a smaller stake.

03 /Opportunities

The key takeaway: AI has made it possible to scale by multiplying the number of products rather than just growing one product larger. This ranges from building similar products for different niches – where AI handles most of the adaptation – to building entirely different products across different domains, all sharing the same underlying AI platform.

A personal caveat: complete delegation of product development to AI, without a strong human operator leading each product, is unlikely to produce great results.

That's why Infinity Constellation putting an experienced human founder at the center of each company matters. That's the principled difference from the corporate startup accelerator model, which historically has a poor track record.

The new model works because the parent company offers something genuinely useful beyond money and generic business advice: concrete, proven approaches to building AI-native businesses, and specific AI platforms that dramatically reduce the time and cost to launch.

So the direction breaks into two streams ⇑

First: treat every internal AI platform and tool as a product – something others could use as a foundation or a building block for new companies. Package it deliberately, not as a side project.

Second: continuously watch the market and accumulate promising product ideas – but resist the urge to launch them all internally as "new directions" The portfolio model works precisely because each idea gets its own operator.

And third: watch for experienced, hungry entrepreneurs nearby – people you can offer your internal platforms and tools to, in exchange for building specific products from your idea bank. Without trying to micromanage them. But also without entertaining anyone who thinks they can "build any platform in two weeks on Claude" You, in turn, need to be offering something more substantive than a generic toolkit.

The AI-native product market is still the Wild West. There's territory to claim – if you approach it with the right model.