Argentina has no AI law and, under President Javier Milei, does not want one. Its AI strategy is a tax and legal-stability regime, the Súper RIGI (Regime of Incentives for Large Investments in New Industries). In early September 2026 the ruling coalition accepted changes to it to win Senate support. Those changes show the trade-offs of building an AI strategy this way.
What the bill does
On June 24, 2026, the Chamber of Deputies approved the bill by 130 votes to 106, with 7 abstentions, according to the chamber's own press office. According to La Nación's explainer, projects must invest at least US$1 billion, with at least 20% committed in the first two years. In return they get a 15% corporate income tax rate, exemptions from import and export duties, gradual freedom to keep export revenue abroad, and 30 years of regulatory stability with international arbitration. That is a deliberate copy of the original RIGI, which the government's announcement presented as a lower-rate, faster-depreciation version aimed at new industries.
The case for the Senate's changes
The strongest argument for the amendments deserves a fair statement. A data-center campus draws power on the scale of a small city. Argentina's grid has a history of stress, and provinces fear that a US$1 billion-plus facility could push residential and industrial users into outages. Allied blocs (UCR, PRO and provincial parties) also have a reasonable political point. A regime that cuts taxes to 15% and offers 30 years of stability should show something to voters beyond capital expenditure that mostly buys imported servers.
The Senate amendments answer those concerns. Per elDiarioAR, data-center investors must build their own electrical substations to reduce supply-interruption risk. According to Parlamentario, companies must also commit to contracting local goods, services and works equivalent to at least 20% of the investment. This applies where price and quality are competitive. Allied blocs pushed to make the percentage count only goods actually produced in Argentina.
Where the design creates friction
The first change is defensible. Making large loads pay for their own grid connection is a common condition, and it protects other ratepayers without banning anything.
The second is harder to defend for AI infrastructure. Frontier data centers run on accelerators, networking gear and cooling systems that no Argentine manufacturer makes at scale. A local-content floor, particularly one tightened to count only domestically produced goods, risks becoming either a paper exercise or a real cost on projects whose economics already depend on tax terms. The competitive-price and quality proviso helps, but it invites disputes over who decides what counts as competitive. Investors price that uncertainty.
Process cost matters too. Because the Senate text differs from the Deputies' version, the bill would have to go back to the Chamber of Deputies if the Senate approves it with changes, and the committee ruling was itself blocked on September 2 after missing an August 26 target. Each round keeps the regime unavailable.
The flagship project is still a stated intention
The cost of waiting is visible in the project that was meant to anchor the strategy. OpenAI and Sur Energy announced 'Stargate Argentina' in October 2025: a roughly US$25 billion, 500MW data center. According to the Buenos Aires Times, as of May 2026 it had not been formally presented to the government, and the RIGI Observatory did not list it. The same report says Sur Energy's co-founder acknowledged the project depends on the RIGI framework being in place. Press reports through September 19 say the large AI data-center plans remain stalled while the Súper RIGI awaits approval.
That means a strategy built on incentives has made its own legislative timetable the binding constraint. A tax regime cannot pull in capital that is waiting for the tax regime. The original RIGI's 30-year stability promise only has value once it is law, and every amendment round pushes back the date when it can be tested.
What proportionate policy would look like
The Milei government's choice not to regulate AI itself is defensible, and it is consistent with evidence that early, prescriptive AI rules can lock in compliance costs before risks are understood. But an absence of AI regulation is not an AI strategy. Three adjustments would keep the Súper RIGI's pro-investment logic while answering the Senate's concerns:
- Keep the substation rule, and pair it with a fast permitting path. Self-built connections should be matched by predictable grid-approval timelines, so the obligation is a cost rather than an open-ended risk.
- Define local content by outcome, not just by percentage. Counting construction, installation, power, and operations services, where Argentine firms compete, would let the 20% floor be met without pretending the country can supply chips.
- Publish the implementing regulations early. Investors have shown they will wait for law, but they will not wait indefinitely for the rules that follow it.
Argentina has a real comparative advantage in energy, cold climate in the south and skilled software talent. The Senate is right to ask that a large data center leave something behind besides a fence and a power line. It should not ask in a way that makes the investment less likely to arrive at all.