Where will the next decade of compute-intensive industry get built? Part of that answer is settled by grid mix: a data hall, a pilot plant and an AI cluster all go where electricity is abundant, clean and available to run flat out. On that test Latin America is one of the strongest regions on earth, and this month the evidence got fresher. Ember’s Global Electricity Review 2026, published in April with 2025 data, puts Latin America and the Caribbean at 65% clean electricity against a 41% global average, with fossil generation down to 35% from 47% in 2015. That 65% is what our September 2025 report printed, and it holds unchanged.

A Grid That Is Already 65% Clean

Composition matters as much as the headline. Hydro is 40% of the region’s clean power in Ember’s breakdown and wind plus solar 19%, so the clean share rests on a large hydro base whose output follows hydrology, with a variable wind and solar layer on top. One note before quoting the number: Ember’s category is clean rather than renewable because it includes nuclear, and 65% is a share of annual generation rather than firm capacity at a particular substation, which makes it the right figure for shortlisting a country and the wrong one for signing a power contract.

Our report’s foreword calls the region’s clean-energy potential critical for scaling compute-intensive industries. We wrote that as a thesis about comparative advantage rather than as a measured finding, and what follows is what already stands behind it.

The region commands world-class biodiversity, booming renewable-energy corridors, and a growing cadre of STEM researchers eager to found science-based companies.

The Ventures Already Trading on That Power

The clean grid is not an abstraction to the companies in our own mapping. Splight, the Chilean venture profiled in the report at TRL 9, sells grid-operations AI that attacks curtailment and congestion so operators can integrate more renewable output. Its card reports USD 26.1 million raised in total, including a USD 12 million seed led by noa, formerly A/O, with EDP Ventures, Elewit, Draper Cygnus, Ascent Energy Ventures and the UC Berkeley Foundation participating. Both figures are extracted from the report and not yet re-verified, so carry them as indicative of scale.

Photio, out of Chile, appears on the IDB’s Deep Tech: The New Wave mapping with nanoparticle additives that turn urban surfaces into photocatalytic air-purification agents, backed by Start-Up Chile and CORFO. Strong by Form, founded in Santiago in 2018 and at TRL 7, uses bio-inspired digital manufacturing to make structural wood components that replace steel, concrete and aluminium, and won its track in Hello Tomorrow’s 2022 Challenge for Sustainable Construction and Infrastructure. Its funding appears in our report in two forms, USD 6.2 million in total on the company card and a EUR 4.8 million seed led by CMPC Ventures in 2023 in the corporate-venture chapter, so keep both rather than pick one.

Sistema.bio, out of Mexico, converts organic waste into clean energy and fertiliser, runs major hubs in Mexico and Colombia with partnerships in a dozen LATAM countries, and raised USD 22.8 million across three Series B rounds between 2024 and 2025 from KawiSafi, AXA Investment Managers, EcoEnterprises Fund and Novastar Ventures, again on figures extracted from the report and not yet re-verified.

None of this is a niche. On the Sling Hub split our report carries for 2024, which we have not re-verified, energy took 13% of total tech funding in the region against 55% for fintech and 6% for deep tech, making energy the second-largest destination for tech capital there. Specialist capital is forming behind it. Draper Cygnus, one of Splight’s investors, launched a USD 50 million early-stage fund in 2025 for deep tech with Latino founders across biotech, space, energy, semiconductors, AI and climate, and GRIDX Fund II came in at USD 30 million with IDB Lab as a USD 3 million LP and a mandate for up to 75 startups.

Shared Facilities Are the Second Half of the Advantage

From quantum rigs and petabyte-scale AI clusters to secure blockchain nodes and biotech wet-labs, Latin American innovators converge on one persistent bottleneck: the region’s limited supply of specialised, capital-intensive R&D infrastructure.

That sentence from our report names four kinds of capital-intensive infrastructure, and the region has moved fastest on one of them. Brazil now counts 11 active biotech hubs, Chile at least three, Uruguay at least seven and Argentina at least one, on figures the report takes from Endeavour’s Brazil Biotech Report and Uruguay Innovation Hub that we have extracted but not yet re-verified. Read them as a measure of biotech wet-lab provision, which is what they count, rather than as a proxy for cluster capacity.

The facilities have names worth knowing. Startuplab.01, launched by CORFO and Fundación Chile with IDB support and a GridX partnership, is a shared lab for deep tech startups working on climate. The Patagonia Biotech Hub opened in Puerto Varas in September 2024, backed by Chile’s Ministry of Science. New Lab, Newlab’s first Latin American hub, sits inside Montevideo’s Innovation Campus with Uruguay’s ANII, Globant and UPM, offering prototyping labs and pilot testbeds. The Parque de Innovación de Buenos Aires spans 12 city blocks and 340,000 square metres.

Our report’s recommendation here is to expand and replicate: double down on open-access labs and pilot plants, copy proven models into new countries and coordinate the investments region-wide, so a venture anywhere in Latin America can reach world-class facilities without prohibitive upfront cost. The same logic extends to compute. A shared cluster placed on a grid that is already 65% clean starts well above the 41% global average on emissions, which is a procurement argument as much as an environmental one.

Compute Is Arriving by Treaty, and Clean Power Is the Bargaining Chip

The largest compute commitments landing in the region are not coming through the venture funnel. Our report’s case study on the United Arab Emirates records Comprehensive Economic Partnership Agreements signed with six LATAM countries, Colombia, Costa Rica, Chile, Mexico, Ecuador and Brazil, more than USD 100 billion committed in targeted investments, and hyperscale data centres launched to meet the region’s surging AI demand, on the Middle East Council on Global Affairs’ account. In Colombia, a CEPA in April 2024 was followed by a Dubai Chambers office in Bogotá and the announcement of three hyperscale data centres to be built by G42, per BNamericas, alongside a 51% year-over-year increase in non-oil bilateral trade. Costa Rica’s CEPA entered into force in February 2024 and has doubled non-oil trade, Chile finalised its accord in July 2024, Mexico entered the framework in June 2025 with a pact elevating semiconductors and AI to top-priority status, and Ecuador followed the month after.

Every figure in that paragraph is extracted from our report and awaiting re-verification, and the section moves fast, so read six countries as a floor. The aggregate points the same way. Our report carries UNCTAD’s finding that developing countries attracted more than USD 530 billion in greenfield digital economy projects between 2020 and 2024, close to 80% of it concentrated in just ten countries, with Brazil and Mexico the only Latin American economies represented, also extracted and not yet re-verified. Two seats at that table is a starting position. What a Latin American government brings to the negotiation is what Ember has just re-confirmed: a grid a hyperscaler can point to when it publishes an emissions number.

The Road Ahead

Which brings us to the number we do not yet have. Our corpus holds no megawatt figure for the region: no data-centre capacity by country, no rack or GPU count, no delivered price for a GPU hour. Three hyperscale data centres is a count of buildings, and we will not turn it into a power envelope by estimating one. Nor is the silence in our record evidence of scarcity on the ground, since demand may already be met by cloud capacity that no deep tech mapping counts. The claim the evidence supports is this: the region’s grid is a real structural advantage for compute-intensive industry, and the capacity side of it is unmeasured, by us and by everyone else.

That is the next dataset, and it is short enough to specify.

Operating and contracted data-centre capacity in MW by country, separating merchant colocation from captive hyperscaler build.

Interconnection queue duration and firm capacity available at candidate industrial sites in the plausible host markets.

Delivered price for GPU-hours from regional cloud availability zones, benchmarked against North American zones on identical instance types.

The contractual terms of the announced foreign builds, specifically whether any capacity is purchasable by locally domiciled research and startup users.

None of that requires new science. The inputs sit in public filings, regulator queues and vendor price lists that nobody has assembled in one place, which is what recommendation 14 in our report proposes: an open-access data commons of standardised statistics, registries and dashboards to cut diligence friction.

For founders choosing where to train, the decisive variable is a delivered price per GPU hour that nobody publishes in aggregate for this region, so ask your vendor in writing. For allocators, energy already ranks second only to fintech among the slices our report names for 2024, and grid-side ventures like Splight sit on the constraint a growing clean share creates. For ministries, the counterparty in a data-centre negotiation knows its own megawatts, and the host should arrive knowing its own. We are building that count. If you hold capacity data, interconnection timetables or regional GPU pricing, write to us, and we will publish what we can verify with the source named.