On 2 September 2026, the Milei government failed to get the Senate's joint committees to sign the dictamen, the committee report that sends a bill to the floor, on the "Súper RIGI." The bill is Argentina's most ambitious attempt to build an AI, semiconductor and data-center industrial base. Allied blocs asked for more changes, including on data-center energy requirements. Diario Panorama reported that the dictamen was "still circulating" without the needed signatures, and that committees would try again the following week. The delay matters for investment, and it also shows an internal tension in the design.
What the bill does
The Súper RIGI builds on the original RIGI (Régimen de Incentivo para Grandes Inversiones), created by Title VII of Ley 27.742, the 2024 Bases Law. Article 173 of that law sets the RIGI's minimum investment in computable assets at US$200 million. The regulation was approved by Decreto 749/2024, published on 23 August 2024.
The new regime raises the threshold to US$1 billion per project. According to DPL News, it cuts income tax to 15% for adhering single-project vehicles and offers 30 years of stability across tax, customs, foreign-exchange and social-security rules. It also allows international arbitration against the state. Eligible activities must be new or experimental in Argentina, which excludes expansions of existing facilities. Parlamentario adds that projects must commit 20% of the amount in computable assets within two years, and must commit at least 20% of supplier payments to local suppliers, where local supply is available at market price and quality. Diputados passed the bill 130–106 in June.
The strongest case for the critics
The skeptics have a serious argument. Thirty-year stability clauses and international arbitration bind future governments on tax and regulation across several electoral cycles, and Argentina has a long record of policy reversals. Opponents also ask why a handful of very large foreign projects deserve better terms than domestic firms. Provincial governors have a concrete worry too: a hyperscale data center can draw as much power as a mid-sized city. If the grid cannot supply it, residents and local industry bear the cost. These concerns are legitimate and should not be dismissed as obstruction.
Where the design undercuts itself
The amendments now being negotiated, though, show the opposite problem: the bill risks asking for too much from the investors it is trying to attract.
Start with local content. The Diputados text fixed 20% of investment for purchases in Argentina. In the Senate, according to Diario Panorama, PRO, UCR and Provincias Unidas pushed for local purchasing even where the domestic product costs up to 15% more than the import. They also asked to raise to 40% the share of tradable goods sourced from local developers. Baenegocios reports that the government accepted limiting the 20% commitment to goods produced in Argentina.
This is a poor fit for the sectors named in the bill. Semiconductors and AI data centers depend on imported accelerators, networking gear and specialised equipment that no Argentine producer makes. A rule that pays a price premium for domestic substitutes taxes the project, and in the worst case it makes capex unviable. The existing text qualifies the obligation with the phrase "at market price and quality," which acts as a sensible safety valve. A 15% price tolerance would weaken it. Local-content rules in other economies have repeatedly raised project costs more than they built domestic capacity, and an AI build-out is the wrong place to repeat that experiment.
The energy dispute is more nuanced. Baenegocios reports that allies and provinces want energy-intensive industries, including AI mega-data centers, to build their own generation. The government resists, arguing this would make the regime less attractive. The compromise reported by Parlamentario and Diario Panorama is narrower: large data centers must build their own electrical substations when their consumption could affect the regional or municipal energy system. That is a reasonable cost-causation principle. Whoever imposes a grid burden should fund the connection. Requiring on-site generation, however, is a heavier demand, because it converts a tax incentive into a power-plant mandate. The better path is transparent grid-connection charges, with developers free to contract for power from the market.
The procedural cost
There is also a process cost. Because the Senate is amending the bill, it must return to Diputados, where the first vote was far from comfortable. Each round adds months for a regime whose pitch is certainty. The original RIGI shows the stakes: the Bases Law set a defined window and threshold, and the government has since modified the implementing rules by decree. Investors weighing a US$1 billion data center commitment will price in legislative churn alongside the headline 15% tax rate.
I could confirm reporting only through early September. I could not verify whether the committees later signed the dictamen or whether the bill reached the Senate floor.
What a proportionate version looks like
- Keep the headline guarantees. Long-term stability is the product being sold, and it is what distinguishes the regime from ordinary tax policy.
- Make local-content duties outcome-based. Reward supplier development, training and technology transfer, rather than mandating purchases at a price premium.
- Charge for grid impact directly. Substation and connection costs on developers are defensible. Generation mandates are not.
- Publish the rules once. Amend decisively in the Senate, then pass quickly in Diputados, so the final text is not renegotiated by decree.
Argentina has real assets for AI infrastructure: cool southern climates, renewable potential and a skilled workforce. The Súper RIGI can convert them into investment, but only if its conditions cost investors less than its incentives give them.