On May 12, 2026, Humberto Barbato, executive president of Abinee — Brazil's electrical and electronics industry association — arrived at the Vice-Presidency office in Brasília with executives from Cisco, Dell, Flextronics, Foxconn, HPE, Lenovo and Positivo. Their target was a single stalled bill. They met Vice President Geraldo Alckmin and Development Minister Marcio Elias to press for a Senate vote on PL 278/2026, the ReData special tax regime for data centres, which had been frozen since late February. Alckmin committed to raising it with Senate President Davi Alcolumbre (Abinee).
The lobbying is understandable. But the episode exposes two separate problems — one political, one structural — that the rush to "unblock" the bill risks blurring together.
What ReData actually does
Authored by deputy José Guimarães (PT-CE) and introduced on February 4, 2026, PL 278/2026 modifies Law 11.196/2005 to create a special regime suspending federal taxes — PIS/Cofins, the IPI industrial tax, and import duties — on equipment used to build or expand data centres in Brazil. The Chamber of Deputies approved it on February 24 and sent it to the Senate the next day (Câmara dos Deputados). The waiver is sizeable: roughly R$7 billion over three years, with R$5.2 billion already baked into the 2026 budget (Capacity; BNamericas).
The incentives come with strings. Beneficiaries must use 100% renewable or clean energy, invest 2% of the value of incentivised equipment in domestic R&D, and — the clause that matters most here — make at least 10% of their processing, storage and data-management capacity available to the Brazilian domestic market (Capacity).
The political blockage is not about merit
First, the easy part. ReData is not stuck because senators doubt its substance. As regulatory analyst Erich Decat put it, "it's a political issue, not a matter of merit." Alcolumbre pulled the bill from the agenda amid friction with the Executive — a retaliatory move, not a policy objection (BNamericas). The underlying provisional measure, MP 1.318/2025, lapsed when the Senate failed to vote, leaving promised tax reductions in legal limbo. That uncertainty is corrosive: capital allocators deciding where to site multi-billion-real AI infrastructure cannot price a regime that exists only as an expired decree. On this narrow point, Abinee is right. Procedural hostage-taking over a budget-neutral, already-funded incentive is bad governance, and the Senate should schedule the vote.
The 10% rule confuses geography with control
The harder question is whether the bill, as drafted, is good policy — and here the domestic-capacity reservation deserves scrutiny rather than a fast-track signature.
The strongest case for the 10% clause is real and worth stating plainly. Civil-society researchers at IP.rec argue that "broad and immediate tax incentives" without binding obligations risk turning data centres into enclaves — highly automated facilities that consume Brazilian energy, land and water while concentrating the gains elsewhere and generating few permanent jobs (IP.rec). Reserving capacity for domestic users is a way of insisting the public gets something back for the foregone tax revenue. That is a legitimate worry, and the renewable-energy and R&D conditions are sensible expressions of it.
But the capacity-reservation clause is the wrong tool. It is sold as a sovereignty measure, yet it conflates where servers physically sit with who actually controls the data and the compute. IP.rec itself makes the distinction the bill misses: sovereignty "cannot be reduced to the physical hosting of infrastructure" and requires strategic control and domestic enterprise — "elements absent from the current proposal" (IP.rec).
The market reality underlines the point. Brazil already accounts for roughly 40% of Latin American data-centre investment, and its São Paulo regions are anchored by the same three foreign hyperscalers — AWS, Microsoft and Google — that dominate cloud globally (Capacity). A 10% domestic-capacity carve-out inside a foreign-operated facility does not transfer control of data, software stacks or AI models to Brazilian hands. It simply requires that a slice of foreign-owned racks be rented to local customers — something competitive markets already do. The clause functions as a soft data-localisation lever dressed as industrial policy: it raises compliance friction and audit cost for operators while delivering little of the autonomy it promises.
A more proportionate design
Proportionate regulation would decouple the two goals. If the aim is fiscal fairness, condition the tax break on transparent, measurable outputs — energy sourcing, water efficiency, the 2% R&D spend, and regional siting bonuses (the bill already cuts requirements 20% for projects in the North, Northeast and Midwest). If the aim is genuine digital sovereignty, the levers are interoperability mandates, data-portability rights, public-sector procurement of domestic cloud, and investment in Brazilian AI capability — not a percentage quota on rack space.
The risk of bundling a localisation quota into an otherwise pro-investment regime is that it invites reciprocal treatment abroad and signals to the very hyperscalers Brazil is courting that capacity will be politically apportioned. Brazil wants to be Latin America's compute hub. The way to get there is a clean, stable, predictable incentive — passed promptly, and stripped of a sovereignty fig-leaf that confuses the address of a server with command over what runs on it.