How many deep tech companies does Latin America have, and which country leads? Ask two credible mappings of the region and you get two different answers, and both of them are right. In this post we take Latin America’s deep tech league tables apart: what happens to the regional map when artificial intelligence is broken out as a vertical of its own, which companies the wider net catches, and the checks a fund, a founder or a ministry should run before trusting any ranking of the region.

Three Counts of One Region, All Defensible

Three mappings of Latin American deep tech are in circulation, and they were never trying to be the same size.

The IDB’s Deep Tech: The New Wave mapped 340 deep tech startups in Latin America in 2023, on a twelve-vertical framework. We carry that count from the IDB rather than having re-derived it ourselves.

EMERGE and Cubo Itaú’s Radar Deep Tech LATAM 2025, presented on 11 September 2025, counts 1,316 companies across the region.

Our database, with a March 2025 data cutoff and sourced primarily from Tracxn, comprises 2,566 companies, filtered down from an initial set of more than 5,000.

The distance between those numbers is definitional. Our report names the decision in one sentence, printed on the same page as the counts:

The main discrepancy comes mainly because we expanded the scope of the database to AI and cryptography, 2 verticals that have significantly grown in the last 2 years.

Our framework is the IDB’s twelve verticals with artificial intelligence and cryptography broken out as verticals in their own right. EMERGE’s narrower frame preserves comparability with the IDB’s 2023 baseline, which is a defensible thing to want. Every count carries its taxonomy and its cutoff date, and ours carries one more caveat worth stating plainly: the 2,566 rests on a proprietary database we cannot republish, so read it as LADP’s count rather than one an outside reader can rebuild. Apply that same test to any mapping you rely on, including the ones that flatter your thesis.

What Breaking Out AI Reveals About the Region

On the IDB’s 2023 mapping, biotechnology accounted for 61% of Latin America’s deep tech companies and artificial intelligence for 11%, the two together at 72%, followed by nanotechnology at 6%, cleantech at 5%, and spacetech and advanced mobility at 4% each. Those are 2023 figures from the IDB and we present them with that date. We profiled that AI cohort in this space in January.

We broke AI and cryptography out because both verticals grew sharply in the years before our cutoff, and folding them into a general software bucket loses information about where founders in the region are actually building. Consider what the wider net catches. Auth0, founded in Argentina in 2013 on standards-based identity and authentication, had raised more than US$210 million by mid-2020, culminating in a US$120 million Series E at a US$1.92 billion valuation, before Okta acquired it in 2021, a landmark regional exit that our report classifies under cryptography. NotCo, in Chile, applies AI to plant-based food formulation and is on EMERGE’s Radar the country’s largest single deep tech recipient by cumulative funding, at roughly US$466 million, about 75% of Chile’s country total.

Splight, also Chilean, sells grid-operations AI that tackles curtailment and congestion to raise renewable integration, and raised a US$12 million seed led by noa alongside EDP Ventures and the UC Berkeley Foundation. Tractian, the Brazilian industrial AI company, integrates hardware, software and AI for predictive maintenance and closed a US$120 million Series C led by Sapphire Ventures, with General Catalyst and Y Combinator among its backers. File that kind of company under software and the region does not lose Brazil. It loses Argentina, Chile and Mexico.

The Countries the Wider Net Finds

Our report puts Brazil at 1,048 deep tech startups, about 40.8% of the database, ahead of Mexico at 358 (14%), Argentina at 256 (10%), Chile at 251 (9.8%) and Colombia at 219 (8.5%). By our own arithmetic those five hold 83% of the companies we map, with a long tail of emerging hubs behind them in Panama, Peru, Puerto Rico, Uruguay and Ecuador, which is where an investor hunting earlier-stage and less crowded markets should be reading.

EMERGE’s Radar orders the same region differently: Brazil at 952 companies, 72.3% of its count, then Argentina at 145, Chile at 72 to 73, Mexico fourth at 53, Colombia at 44 and Uruguay at 21. Both sets of country figures are reported from the two mappings rather than independently re-derived by us. Brazil holds 72.3% of the companies on EMERGE’s mapping, which counts AI ventures inside deep tech, and 40.8% on ours, which breaks AI and cryptography out separately, so the two figures answer different questions rather than competing to answer one.

Read both as company counts and nothing else. Neither is a share of capital, and the two are not proxies for one another. On the same Radar, private deep tech investment in 2024 ran Chile US$607 million, Argentina US$486 million and Brazil US$216 million, and EMERGE names brain4care, the Brazilian non-invasive intracranial-pressure monitor, its largest domestic deep tech recipient of that year.

Cities move too. Our report has São Paulo first with 329 startups, roughly 29% of all ventures in the top-ten cities, with Buenos Aires, Santiago and Mexico City forming a strong second tier, Bogotá anchoring the Andean corridor, and Brazil fielding five of the top-ten hubs: São Paulo plus Rio de Janeiro, Curitiba, Belo Horizonte and Porto Alegre. EMERGE counts 467 companies in São Paulo state. A city and a state are different geographies, so those two numbers should never be set against each other.

Four Checks Before You Trust a Ranking

None of this makes regional rankings useless. It makes them readable, provided you ask four questions first.

Which vertical list produced it. A table built on the IDB’s twelve verticals and a table that breaks artificial intelligence and cryptography out on top of them describe different populations, and the ranking below the leader is where they diverge most.

Whether it counts companies or capital. Company counts tell you where ventures are formed. Capital tells you where they are funded. In this region those two maps do not sit on top of each other.

Which geography and which vintage. City or state, country or region, and a March 2025 cutoff read against a September 2025 presentation.

Whether anyone outside the publisher can rebuild it. Ours cannot be rebuilt from anything we are able to publish, and we would rather say so on the page than bury it in a footnote.

The same discipline applies to funding totals and investor counts. Dealroom put Latin American deep tech funding at US$138 million in 2024, and Sling Hub put it at US$536 million for the same year, on a narrower and a broader definition, the broader one including debt instruments. The IDB counted 65 VC funds with at least one deep tech investment in the region in 2023, while Hello Tomorrow’s 2025 investor mapping identified nearly 40 funds active there, on different inclusion rules, so that pair is not evidence of investors leaving. All four figures are extracted from those sources and carry their dates; we have not re-derived any of them.

The Work Ahead: One Regional Dataset, Openly Versioned

The honest position is that the region does not yet have a shared evidence base, and no single institution should be the one to declare a winner. Our report says what the fix looks like: a regional, open, versioned and auditable dataset with a shared taxonomy, transparent inclusion criteria, and a public correction and contribution workflow, ideally API-interoperable and stewarded by multi-stakeholder governance. That is recommendation 14 of the twenty we published, an open data commons paired with a unified grants portal so founders and officials can compare funding types, ticket sizes, eligibility and deadlines in one place.

This matters most to the people the tables leave out. A founder building an AI company in Mexico City sits inside one regional dataset and outside another, and that classification decides which reports, ecosystem maps and public programmes ever find her. A shared taxonomy is market infrastructure, and it is cheap relative to what it unlocks.

What to Do With This

For investors: establish which taxonomy produced a league table before you trust its rank order, and check that a mandate written around deep tech is aiming at the same population its underlying data counts. A fund quoting country rankings from one mapping while screening against a different vertical definition is comparing two populations without knowing it.

For founders: whether your company appears in these counts is a classification outcome, and it is worth knowing which frame you fall inside before you approach a programme or a fund. Ask the question in the first meeting.

For ministries and development banks: if a target is going to be set against a company count, publish the taxonomy alongside the target, and prefer a count someone outside your institution can reproduce.

We are building toward that shared dataset, and we would rather build it with the region than for it. If you maintain a mapping, run a programme, or think our vertical list draws its lines in the wrong places, write to us. Read Accelerating Deep Tech in Latin America for the full country and city tables, and tell us where your count differs from ours and why.