On Tuesday, October 6, Brazil’s grid operator did something no system operator has done before at this scale: it sold scarce transmission access to the highest bidders instead of handing it out in queue order. The Operador Nacional do Sistema Elétrico (ONS) offered 228 megawatts of leftover transmission capacity — margem de escoamento — across four connection points in São Paulo state. Seven projects bid for 451 megawatts, nearly twice what was available. When the process closed, every megawatt had gone to three data centers: Pahlavan One, Pahlavan Three, and Digital DC Brasil.
For investors, the headline number matters less than the precedent. Grid access — the interconnection slot this desk has called a gating document on par with the PPA itself — just became a priced, competitive input. The waiting line is now an auction.
What PNAST actually changed
The allocation ran under Brazil’s Política Nacional de Acesso ao Sistema de Transmissão (PNAST), the National Policy for Access to the Transmission System, established by federal decree in 2025. Before PNAST, access to Brazil’s basic grid (rede básica) worked the way interconnection works in most of the world: first come, first served. File your request, join the queue, wait your turn.
PNAST replaces the queue with “access seasons” — batch windows, at least two per year starting in 2027, in which ONS analyzes all requests together and allocates capacity on technical and competitive criteria. Competition only kicks in where it has to: at the specific buses where requested capacity exceeds available flow room. Where there is no congestion, access stays routine. Where there is, the scarce slots go through a competitive process — which means, in practice, to whoever values them most and can pay.
Lawyers following the policy told BNamericas the framework brings clearer rules and should reduce legal risk in the sector — a meaningful statement in a country where grid-access disputes have a long litigation tail. Projects with access opinions already issued, and requests filed before the decree, were grandfathered; the new rules bite on new requests and on expansions of contracted transmission usage (MUST, the montante de uso do sistema de transmissão).
Why São Paulo, why now
The first season’s numbers explain the urgency. ONS logged 367 registrations from generation and consumption projects totaling more than 20 gigawatts of access requests across 64 buses of the National Interconnected System. Counting the transitional phase, the operator analyzed over 43 gigawatts of requests — and found room for about 30.7. The rest hit the physical limits of the wires.
Two forces are colliding. On the supply side, Brazil’s renewable buildout keeps filing interconnection requests. On the demand side, data centers are arriving fast, pulled by the Redata tax-incentive regime for digital infrastructure. Industry executives are already warning that data-center growth could outrun transmission expansion and become the binding constraint on new projects. São Paulo — Brazil’s economic core, where load is densest and spare flow room scarcest — was the natural place for the first competitive allocation. ONS and the Ministry of Mines and Energy presented the season’s results in Brasília on October 7, the day after the process closed.
The mechanism, and where the money goes
Here is the part investors should study, because it is the design choice that makes the whole thing politically viable: the proceeds from competitive allocations do not disappear into general revenue. The government directs them to modicidade tarifária — tariff moderation, the bundle of measures used to hold down electricity costs for Brazilian consumers.
That single decision defuses the obvious objection. Auctioning grid access looks, at first glance, like a toll on connecting to the grid — a charge that raises costs for exactly the investment the country wants. Routing the proceeds into tariff relief converts the scarcity rent into a consumer benefit, which is what lets the regulator price a bottleneck without being accused of manufacturing one. Compare that to the cost-allocation fights around U.S. data-center load, where the question of who pays for grid upgrades is litigated state by state with no clean answer. Brazil’s answer is elegant: the winners pay, and households get the rebate.
Reading the result as an investor
Start with what the bidding revealed. Seven projects chasing 451 megawatts against 228 on offer is roughly a 2-to-1 oversubscription ratio — the market’s way of saying the clearing value of access at those four São Paulo buses is well above zero. And the winners were all data centers. Not generators, not green hydrogen — the loads. That tells you where the willingness to pay sits: with the customers whose projects die without firm grid access, and whose economics can absorb an access premium because the alternative — waiting years in a queue, or building behind the meter at a steep premium — costs more.
This is the same logic this desk laid out in the data-center-power asset class piece: interconnection is the critical path, and speed has a price the buyers can pay. Brazil just gave that price a formal market.
Second, notice what got de-risked. A data center holding competitively allocated access at a São Paulo bus now owns something bankable: a firm, system-operator-awarded grid position. In project finance terms, that is a milestone on par with a signed PPA or a completed interconnection study — it converts development risk into a secured input. Lenders underwriting these projects can point to allocated access the way they would point to any other gating document. The auction did not just allocate capacity; it manufactured certainty, which is the raw material of leverage.
The new line item in project economics
Every project finance model for a Brazilian large-load project now needs a new line: the cost of access. Consider, for illustration, how it flows through the structure. Access payments are an upfront or contracted cost of the SPV — they sit alongside development spend, senior to debt, and they must be covered by the revenue contract or the sponsor’s equity. If access clears at meaningful prices in congested zones, it does three things: it raises the equity check at financial close, it tightens the DSCR by adding a cost the offtake price must cover, and — this is the subtle one — it rewards sponsors who can underwrite regulatory process risk, because winning access becomes a competitive advantage in itself.
That last point is where the smart money goes. In a world where access is auctioned, the scarce skill is not just building cheap — it is winning the right to connect. Developers with local regulatory expertise, balance sheets to back their bids, and portfolios that let them compete across multiple buses have a structural edge over single-project shops. Expect consolidation pressure: the access market favors scale, just as PPA markets did before it.
Who wins, who pays
Map the winners and losers the way you would map any market reform.
Winners. The three data center projects that secured São Paulo access — they converted an uncertain multi-year queue position into a firm allocation. ONS and the system planners, who now get batch visibility into real demand instead of a speculative queue — 43 gigawatts analyzed is planning data you cannot get from a first-come line. Brazilian consumers, to the extent auction proceeds actually flow into tariff moderation. And, over time, transmission developers: visible, priced scarcity is the clearest possible investment signal for where the grid needs reinforcement.
Losers, or at least payers. The four bidding projects that walked away empty-handed — they spent time and money qualifying 451 megawatts of demand and got nothing, which is the honest cost of competition. Smaller developers without the balance sheet to compete in price-based allocation, who may find the new regime harder than the old queue, where patience substituted for capital. And in a diffuse sense, anyone who preferred the old certainty that filing early guaranteed a place in line.
The bear case
Intellectual honesty requires the other side, and there are three real objections.
First, scale. 228 megawatts against 43 gigawatts of analyzed requests is a pilot — roughly half a percent. One competitive allocation at four buses does not prove the model works at system scale, and the seasons from 2027 will be the actual test. Do not price the thesis on the pilot.
Second, concentration. Price-based allocation systematically favors the deepest pockets. If every congested bus in Brazil clears to hyperscalers and large developers, the access market becomes a moat for incumbents — efficient in the textbook sense, but politically fragile. The tariff-relief recycling helps, but watch for the backlash when a small renewable developer loses a bus to a data center for the third season running.
Third, regime risk. PNAST is a 2025 decree — young, untested, and already the subject of at least one filed dispute. The rules of the seasons, the treatment of expansions, the exact competitive mechanics: all of it can still move. Investing against a regulatory framework in its first year means underwriting the regulator as well as the asset.
Could this travel?
The question every infrastructure investor should be asking: does auctioned access stay a Brazilian experiment, or is it the future of interconnection everywhere?
Start with the United States, where the problem is worst. More than 2,000 gigawatts of generation and storage are waiting for U.S. grid access, and FERC’s Order 2023 reforms — cluster studies, first-ready-first-served, commercial readiness deposits — are all attempts to make the queue smarter without pricing it. The American debate keeps circling the same drain: queue reform changes the order of access, but nobody pays for the scarcity itself. Brazil just showed what the next step looks like.
Would it survive American law and politics? Open-access tradition, FERC jurisdiction, and fifty state commissions make a direct transplant unlikely — imagine the litigation the first time a utility auctioned interconnection rights at a congested substation. But the direction of travel is what matters. The U.S. is already moving toward readiness deposits and withdrawal penalties, which are baby steps toward pricing queue positions. Brazil jumped to the endpoint: explicit competitive pricing of the scarce slot. Expect the idea to surface in American policy debates within a couple of years, dressed up as “efficient allocation of constrained interconnection capacity.”
The deeper pattern is scarcity pricing migrating down the stack. Power markets learned decades ago to price scarce energy through locational marginal pricing, then scarce capacity through capacity markets, then scarce flexibility through ancillary services and scarcity pricing. Grid access was the last unpriced bottleneck — allocated by waiting time, which is just a price paid in delay instead of money. Brazil priced it. That is not a local quirk; it is the system completing its logic.
What to watch
Four things, in order of importance. Clearing prices — ONS has not disclosed what the winners paid, and the price level is the single most informative number in this story: it sets the benchmark for every future season and every project model. The 2027 seasons — two per year, bigger volumes, more buses; that is when we learn whether the pilot scales. Secondary rights — whether allocated access becomes transferable, which would create a genuine secondary market in grid positions, and a new asset for traders. And copycats — which regulator, in which country, proposes the next competitive access regime. The first follower tells you it is a trend.
The bottom line
Grid access was the last free input in power project development — free in money, ruinously expensive in time. Brazil just put a price on it: 228 megawatts, four São Paulo buses, seven bidders, three data centers, one precedent. For investors, the read-through is straightforward. Underwrite access the way you underwrite offtake — as a priced, scarce, bankable input. Favor developers who can win it. Watch the seasons, watch the prices, and watch who copies the model. The queue had a good run. The auction is the future.