On April 1, 2026, between midnight and 12:15 a.m., Belgium’s cross-border marginal aFRR price for reducing power output averaged 132 euros per megawatt-hour, more than four times the 32 euro average price paid for increasing output in the same quarter-hour. Basic merit order logic says that should not happen, and the reason it does reveals a structural feature of how Europe’s automated balancing platform pools bids across an expanding set of national grids.
PICASSO, the Platform for the International Coordination of Automated Frequency Restoration and Stable System Operation, went live in June 2022 under the EU’s Electricity Balancing Guideline, initially connecting only the transmission power system operators of Germany, Austria and the Czech Republic. Participation has broadened steadily since, with a legal deadline forcing full TSO connection by July 2024 and the Baltic operators joining in April 2025 following their earlier accession to PICASSO’s manual reserve counterpart, MARI, in October 2024. The platform’s Activation Optimisation Function clears a shared cross-border merit order every four seconds, producing 225 distinct cross-border marginal prices, one per direction, within each 15-minute market time unit before those prices are aggregated for settlement. That granularity is what makes the Belgian example visible at all: a single quarter-hour containing a clear, sustained price inversion between the two directions rather than a momentary statistical artifact.
The intuitive expectation is that a TSO paying to increase generation, an upward aFRR activation, should generally pay more than it receives for a downward activation that lowers output, since a participant reducing production is typically compensated less than one increasing it to cover the same shortfall. Reverse that relationship and the TSO would, in theory, be able to activate both an upward and a downward bid simultaneously and pocket the difference, a counter-activation that PICASSO’s rules explicitly prohibit, unlike MARI, where the practice is permitted. A related safeguard, the International Grid Control Cooperation platform, separately nets opposing imbalances between neighboring TSOs before aFRR bids are activated at all, specifically to avoid the kind of wasteful simultaneous opposite-direction activation an exploitable price gap might otherwise invite. Both mechanisms assume the price inversion itself is something the market design should prevent a TSO from profiting off, rather than something that should not occur in the first place, which is the more revealing distinction. The inversion is not a flaw the system fails to catch; it is a known, tolerated feature that the counter-activation ban and IGCC’s netting function are built specifically to neutralize.
The mechanism behind the inversion sits in how the merit order lists themselves are constructed. Because PICASSO settles as pay-as-clear, all accepted bids in a given direction clear at a single marginal price rather than at their individually submitted prices, but the submitted bids underlying that clearing price still reflect genuine participant intent. In the French example underlying this pattern, at least one bidder was willing to accept a negative price to increase output, effectively paying to generate more, while other bidders on the downward side demanded more than 100 euros per megawatt-hour to reduce production. Both positions can be individually rational for reasons specific to each asset, a generator that loses more value shutting down and restarting than it would from a period of negative revenue, or an inflexible unit that can only reduce output at a cost tied to its own operating constraints, and pay-as-clear settlement means the resulting marginal prices for the initial megawatts on each side of the merit order can diverge sharply before converging further down the stack. When an uncongested zone spans multiple countries, the shared merit order list pools bids from every connected TSO in that zone, and the more grids feeding into that common list, the higher the likelihood that this kind of price divergence at the margin produces a quarter-hour where the downward price clears above the upward price, exactly the dynamic the source data illustrates for Belgium and France.
That cross-border pooling is also the platform’s core selling point, which sets up a genuine tension the regulatory record has already acknowledged. ACER’s July 2024 decision amending PICASSO’s balancing rules stated plainly that “since the PICASSO platform started operating, some inefficiencies have been observed,” an official acknowledgment that real-world price formation on the platform has not consistently matched its stated purpose of making balancing energy activation cheaper for European consumers. Expanding the pool of connected TSOs, the same expansion that brought the Baltic operators on board in 2025 and that continues to broaden the merit order lists referenced in this analysis, increases the platform’s liquidity and theoretical efficiency, but it also increases the number of distinct bid curves being merged into a single cross-border list, raising the odds that asymmetric bidding behavior on one side of the market produces exactly this kind of directional price inversion. Whether that tradeoff nets out in consumers’ favor is not something a single quarter-hour of Belgian data can answer, but the fact that these overlapping merit orders recur with enough regularity to be systematically tracked through tools like balancing.services suggests the inversion is a structural byproduct of how PICASSO aggregates an expanding, heterogeneous set of national balancing markets rather than an isolated anomaly tied to unusual grid conditions on any single day.
