Argentina is assembling the incentives to host some of the largest AI infrastructure in Latin America, but not the plan to benefit from it. At the nadIA Summit at Universidad Torcuato Di Tella, researchers argued that the country has no national AI strategy, while Brazil and Chile have already funded theirs. The risk they describe is concrete: Argentina becomes a power supplier to foreign data centers while the high-value work, in models, chips, services and research, happens elsewhere.
What was said, and what is actually on the table
The summit was organised by nadIA, the Argentine AI node formed in 2025 by CEPE and Fundar. Cronista's report dates it to 2 October 2026. It quotes Di Tella economist Eduardo Levy Yeyati saying a real strategy needs "concrete plans with funds, accessible data and regulatory sandboxes", not just documents. It also notes that Brazil and Chile have already funded their plans, and that Argentina has made no clear budget commitment.
The backdrop is Stargate Argentina. Sur Energy signed a letter of intent with OpenAI for a Patagonia data center worth up to US$25 billion, according to Chequeado's explainer. That is a letter of intent, not a completed investment. It also arrives alongside the "Súper RIGI", a bill the government sent to Congress in June 2026 to attract very large investments in AI, semiconductors and digital infrastructure.
The case for the government's approach
The strongest argument for the incentive-first approach is that capital is mobile and compute is scarce. Few countries can offer cheap, abundant power and land at scale. A government that waits to finish a strategy document may lose the project to a neighbour. Argentina also learned from its first investment-incentive regime that legal certainty matters: the original RIGI, created by Law 27.742 and regulated by Decree 749/2024, set sector minimums, including US$1 billion for technology, and offered long-term stability to investors wary of Argentine policy swings. On that logic, a data center is a sensible first step, and a strategy can follow.
Why incentives alone do not capture value
That logic has a gap. Chequeado reports that the Súper RIGI sets a US$1 billion minimum per project, guarantees benefits for 30 years, cuts corporate income tax to 15% (against 25% under the original RIGI), and exempts import duties, including on intermediate goods. It also reports that the bill lacks explicit environmental mandates, unlike approaches in neighbouring countries. A project with zero import duties and a flat 10% social security rate for new hires does not by design buy much local supply-chain activity. The bill does not oblige foreign operators to train local engineers, buy local services or share compute.
The summit speakers pointed to Brazil as the contrast. They noted that Brazil requires data center contracts to reserve computing capacity, which turns an infrastructure deal into access for domestic researchers and firms. Levy Yeyati argued that Argentina lacks the scale to negotiate alone and should bargain jointly with other countries in the region: "we won't ever have that scale, we'll have to share resources."
A tax regime that is generous on the way in and silent on local capability is an energy-export contract with a better brand.
Jobs: the missing evidence
The employment worry is real but poorly measured. According to the Cronista report, Daniel Yankelevich of Fundar noted that Argentina's large informal economy, where many people hold several jobs at once, is not captured in the international studies that predict AI's labour effects. Levy Yeyati warned that the likely outcome is not mass unemployment but precarity, with workers moving into less stable, lower-paid roles. Policy cannot respond to a problem it cannot measure, and that is a cheap thing to fix: a statistics office survey costs a fraction of a data center subsidy.
What proportionate policy looks like
This is not an argument for heavy AI regulation. Argentina's strengths are openness and low barriers, and a licensing regime would waste both. What the country lacks is capacity-building, and that can be done without restricting anyone:
- A funded, short strategy. Name the budget, the owner and three or four measurable goals, such as compute access for universities and startups, public data sets, and training.
- Light-touch conditions on incentives. Tie the largest benefits to capacity reservations for local users and workforce programmes, as Brazil does, rather than to bans or mandates on how AI is built.
- Regulatory sandboxes. These were named at the summit as a core element and let firms test products under supervision without full compliance burdens.
- Labour data first. Measure AI's effect on informal and multi-job workers before designing protections.
- Environmental disclosure. The bill has no explicit environmental mandates, so at minimum, publish power and water use for projects drawing on public grids.
The bottom line
Argentina's government is right that it must compete for AI investment, and right to give investors certainty. But certainty for investors is not an industrial strategy. If Congress finishes the Súper RIGI without attaching any plan for local capability, the country may succeed in hosting compute and fail to benefit from it. The fix is not more regulation of AI. It is a funded plan that makes the incentives pay off at home.