Venture Studio Benchmark: The US Track Record and Brazil's Opening

Here is the venture studio benchmark in one line. The model builds companies in-house, hands a founder a team and first capital on day one, and takes co-founder economics in return. That structure has a twenty-five year track record across the US and Europe, and almost none of it is in Brazil yet. Idealab started in 1996. The studios that followed produced IPOs, unicorns, and, by the most-cited industry figure, an internal rate of return near 50% against roughly 19% for traditional venture capital.

This piece is that benchmark read builder by builder, with the numbers that hold up and the ones that do not. Avante Ventures is a venture studio building AI-native companies in Brazil and Latin America, so we read this record as proof and as opening at the same time. The model works. No one is running it AI-native in Brazil. That gap is the entire thesis.

A venture studio, also called a startup studio or company builder, conceives companies in-house and staffs them, rather than investing in founders who arrive with their own. That single design choice is what the rest of this benchmark measures. You can read the longer version of [why the structure fits Brazil](/why-avante).

A model proven by twenty-five years of builders

The venture studio is not a 2020s invention, and that is the first thing the record tells you. The benchmark builders were founded between 1996 and 2015. This is a proven structure with public exits attached to it, not an experiment running on a pitch deck.

What unites them is simple. Each one created companies in-house, with operators inside from week one, rather than writing checks into other people's startups and hoping. The track records are specific and, in most cases, independently verifiable. Read the six together as one claim with six data points. Building on purpose, with the build team and the capital under one roof, produces public exits at a rate that pure check-writing does not.

• Idealab. Founded 1996 by Bill Gross. More than 150 companies created and 45 or more IPOs and acquisitions, per [Caltech](https://board.caltech.edu/board-members/mr-william-t-gross-bs-81) and the firm.
• Rocket Internet. Founded 2007 in Berlin. Over 100 companies built, with Zalando, HelloFresh, and Delivery Hero all reaching public markets.
• eFounders, now Hexa. Founded 2011 in Paris. Roughly 50 companies and about $5B in cumulative value created, per [Hexa](https://www.hexa.com/).
• Human Ventures. Founded 2014 in New York. 60 or more companies and three unicorns, per [Fortune](https://fortune.com/2024/04/15/human-ventures-approaches-50-million-raised-second-fund/).
• Pioneer Square Labs. Founded 2015 in Seattle. More than 35 companies spun out, including Boundless and Recurrent, per [Ascend VC](https://www.ascend.vc/blog/tag/Pioneer+Square+Labs).
• Atomic. Founded 2012. Built Hims and Hers, now public on the NYSE under HIMS.

What returns does the venture studio model actually generate?

The most-cited number says the studio model returns far more than traditional venture. Per the Global Startup Studio Network, studio-created startups show an average internal rate of return near 50% against roughly 19% for non-studio startups. Avante cites this as studio IRR of ~50% versus an industry-standard ~19% for traditional VC, roughly 2.5x, and always as the GSSN benchmark rather than any single firm's realized return.

Now the caveat that makes the figure usable instead of embarrassing. Those GSSN numbers are self-reported and trace to a single 2020 to 2022 white paper. The largest independent study of the model, [Big Venture Studio Research](https://inniches.com/big-venture-studio-research) published in December 2024, does not reproduce that magnitude. Read the absolute IRR as directional, not as a promise, and never as a guarantee of any one studio's outcome.

What survives scrutiny is the scale and the mechanism. Enhance Ventures counted more than 560 studios operating worldwide, so the model has become a category of its own. A studio [outperforms for structural reasons](/library/why-venture-studios-win-latam) you can name. Plumbing solved once. Operators in the unit-economics model in the first weeks. A repeatable system that compounds across ventures. Argue with the exact multiple if you want. The direction of it holds. That is the honest read of the studio benchmark, worth far more as a working mechanism than as a single headline percentage.

Are there venture studios in Brazil yet?

Here is the part of the benchmark that should interest a Brazilian founder most. None of it is Brazilian. The model that produced Zalando and Hims was built in Berlin, Paris, New York, and Seattle, and the category has barely arrived in Brazil.

The names most often called Brazilian studios are not studios. [Domo Invest](https://domo.vc/), to take the most cited example, is a traditional venture capital fund rather than a company builder. There is no benchmark AI-native venture studio in Brazil yet. The lane is open, and it is open for a structural reason rather than a lack of talent.

• The surface is large. Services account for roughly 70% of Brazilian GDP, per IBGE, with low software penetration.
• The buyers are under-digitized businesses that domain operators understand and generalist VCs usually do not.
• The competition for an AI-native, local studio is close to zero.

The AI-native edge the benchmarks never had

Every studio in the benchmark shares one quiet limitation. They were all designed before AI could build. Idealab in 1996, Rocket in 2007, even Pioneer Square in 2015, all assume a large engineering team and a Series A to reach scale, because in their era that was the only way to do it.

That assumption is now wrong, and the cost curve is the reason. AI infrastructure is now cheap enough to deploy without a Series A. A studio venture launches 6-9 months ahead of a comparably funded standalone team, and solving company plumbing once routes roughly $300K-500K of effective capital per venture into product and traction rather than overhead.

Be honest about the limit. Cheap AI is not a moat by itself, because it lowers the barrier for competitors at the same time. The edge is running the proven studio structure with AI agents inside from day one, in a market where no one else is doing it. That combination is exactly what the benchmark builders never had access to.

How Avante runs the playbook

Avante Ventures is a venture studio building AI-native companies in Brazil and Latin America, and the benchmark is why the model fits here rather than a hope that it might. The structure is proven abroad. The market is open at home. The cost curve finally allows the build to start lean.

In practice that means Avante launches 3-4 ventures per year through a six-stage system. Research, Partner, Build, Traction, Revenue, Compound. It deploys $500K-1.5M per venture across pre-seed and retains co-founder economics, with operating partners staying engaged through the first revenue milestone. The recurring pattern is the [copilot to data to fund flywheel](/library/copilot-to-data-to-fund-flywheel). Build an AI copilot to generate proprietary data, then use that data to raise and deploy capital. You can see [how the studio operates](/principles).

The current portfolio follows that pattern. Alphajuri in judicial assets, WIR in insurance pricing, and BR Auction Intel in real estate auctions. The benchmark says the studio model produces IPOs and unicorns over twenty-five years of patient building. The opening says Brazil has none of it yet, not AI-native. Avante is building from that gap rather than toward it.