If we translate the CBAM problem into something an industrial CFO can actually act on, the question is not “Will CBAM matter?” The question is “How many terawatt-hours of verifiable low-carbon electricity attributes do we need to defend EU sales, what is the cheapest credible way to secure them, and what happens to economics if the grid is late or curtailment rises?” The fastest lever is electricity procurement because it can be contracted well before process equipment is rebuilt, but it only works if the procurement is designed as a portfolio with firming and balancing rather than as a collection of standalone PPAs that fail under congestion.
A practical way to structure this is to build a Serbia “CBAM buyer map” anchored around the five initial CBAM categories that touch Serbian heavy industry most directly: iron and steel, aluminium, fertilisers, cement, and electricity exports. Each has a different electricity intensity and a different willingness to pay for “green attributes” depending on how exposed it is to EU buyers, how tight its margins are, and how much of its embedded emissions can realistically be reduced through power procurement versus process change.
Start with the energy-intensive buyers inside Serbia’s industrial base. In steel, the relevant procurement logic differs by route. For an integrated steel route, electricity does not dominate direct emissions, but it still matters commercially because EU buyers increasingly treat renewable procurement as a gating criterion and because indirect emissions reductions are a visible early step. In aluminium rolling and extrusion, electricity and heat inputs are far more central to unit costs and to the emissions narrative. In fertilisers, electricity matters both directly and through gas-linked cost structures, with the export frame under CBAM pushing buyers toward more transparent emissions accounting and a stronger preference for credible decarbonisation pathways. Cement is electricity-intensive but also process-emissions-heavy; green power helps, but it is not the whole solution. Electricity exports are the simplest in accounting terms and the most exposed to market structure and interconnection economics.
A financing-grade base case for Serbia’s industrial green power requirement by 2028–2030 is 1.5–2.5 TWh per year of contracted renewable attributes delivered through credible structures. An upside case, if EU customers harden supplier requirements and if Serbia’s exporters pre-emptively lock in green procurement as a competitive shield, is 3.0–4.0 TWh per year. The reason these ranges are realistic is that they do not assume instant full decarbonisation; they assume targeted procurement aimed at reducing reportable indirect emissions and stabilising buyer confidence.
To make this concrete, a workable “buyer map” can be expressed as an indicative allocation of green electricity demand across segments. In a base case of 2.0 TWh per year, a plausible split is roughly 0.6–0.8 TWh for steel and downstream metal fabrication tied to EU supply chains, 0.4–0.6 TWh for aluminium processors and metalworking clusters, 0.3–0.5 TWh for fertiliser and chemicals producers with EU exposure, 0.1–0.2 TWh for cement and building materials with export orientation, and 0.2–0.4 TWh to firm and green-label electricity export tranches or to support utilities/aggregators providing green supply blocks to industry. The point is not the exact decimal; the point is that a small number of buyers can account for a large share of the required volume, which makes aggregation economically dominant.
Once you express the target in TWh, the MW translation becomes unavoidable. If Serbia tries to cover 2.0 TWh per year purely with solar, it typically needs roughly 1,200–1,400 MW of solar nameplate at 17–19% capacity factor, and it immediately creates a synchronized midday volume problem that collapses capture prices and forces curtailment unless storage and exports scale with it. If Serbia covers the same 2.0 TWh primarily with wind, it needs roughly 650–750 MW of onshore wind at 32–38% capacity factor, which is a far smaller MW footprint for the same delivered energy and produces electricity in a less synchronized pattern that the system can absorb with lower curtailment and less price cannibalization. The most bankable structure is not either/or; it is a mixed stack that uses wind as the stability backbone and solar as the volume layer at strong nodes, with storage and aggregation acting as the insurance layer.
A credible base-case supply platform that can deliver ~2.0 TWh per year with manageable system stress looks like 400–500 MW wind plus 400–600 MW solar, supported by 100–200 MW / 200–400 MWh of battery storage and a portfolio-level aggregator that firms delivery blocks. This stack is not chosen for ideology; it is chosen because it reduces the probability that the industrial buyer ends up paying for “green” PPAs while receiving unreliable attribute volumes due to congestion and curtailment.
On CAPEX, the ranges are straightforward in concept and brutal in implication. Utility-scale solar in Serbia and comparable SEE conditions typically lands at €0.55–0.90 million per MW, rising toward €0.95–1.10 million per MW where grid works are heavy. Onshore wind typically lands at €1.20–1.80 million per MW, with dispersion driven by civil works, access, turbine class, and HV distance. Two-hour battery systems commonly land at €0.35–0.55 million per MWh fully installed. Under the mixed base case of, say, 450 MW wind, 500 MW solar, and 150 MW / 300 MWh of batteries, a financing-grade all-in CAPEX envelope typically sits around €1.1–1.8 billion once owner’s costs, interconnection, and realistic contingencies are included. The upside case that pushes toward 3.0–4.0 TWh per year—for example 700 MW wind, 800 MW solar, and 200 MW / 400 MWh—moves the envelope toward €2.0–3.2 billion, and the spread is driven less by turbine and module prices than by grid reinforcement intensity and the cost of building at second-tier nodes.
Returns depend on whether these projects are built as merchant plants or as industrial procurement infrastructure. Under CBAM pressure, the most investable structure is typically a set of long-term contracts—often 10–15 years—with settlement terms that allow the industrial buyer to treat the power and its attributes as a stable input rather than as a volatile trading position. Under those conditions, wind and solar projects in the region typically support unlevered returns in a 7–10% band when cash flows are truly contracted and curtailment is contained. Wind can sustain stronger resilience within that band because its capture price erodes more slowly with penetration and because its curtailment dynamics are more event-driven rather than structurally synchronized like solar. Solar can still hit target returns, but it is far more sensitive to curtailment and to midday price compression.
Curtailment sensitivity is where the CBAM buyer map becomes a financial model rather than a policy memo. If the green supply platform delivers 2.0 TWh per year, each 1% of curtailment equals 20 GWh of lost eligible volume. At an all-in green electricity value of €70–90/MWh, that is €1.4–1.8 million of annual value erosion per curtailment percentage point, and it is not a one-off—it repeats every year. At 2% curtailment, the platform leaks €2.8–3.6 million annually. At 5%, it leaks €7.0–9.0 million. At 10%, it leaks €14–18 million, before you price the second-order damage: the remaining delivered volumes often earn a lower capture price because the system is saturated in the same hours when the renewable output is highest. This is exactly why wind-heavy mixes behave better: wind does not concentrate output into the same narrow window every day, so it does not create the same structural curtailment and capture-price collapse dynamics as solar at scale.
For equity IRR, the conversion is mechanical. In a contracted base case targeting 8–10% unlevered equity IRR for the supply platform, moving from 2% curtailment to 5% typically compresses IRR by roughly 60–120 basis points, depending on debt sizing and whether curtailment is uncompensated. Moving from 2% to 10% can compress IRR by 150–250 basis points, which is often enough to push projects below institutional hurdle rates unless the contracted price is unusually strong or CAPEX is unusually low. Wind-heavy portfolios tend to show smaller IRR erosion for the same curtailment percentage because their base capture price is stronger and because curtailment tends to be localized rather than structural, which reduces the frequency of “chronic” loss-making hours.
Grid integration constraints are the biggest hidden variable in Serbia’s CBAM response because the buyer ultimately cares about delivered, verifiable volumes, not theoretical annual MWh. The first constraint is connection-node saturation: industrial-scale solar tends to cluster at a small number of strong substations, and once those saturate, marginal MW becomes disproportionately expensive and more exposed to curtailment. The second constraint is voltage and reactive-power management, which becomes a recurring owner’s cost at scale. The third constraint is ramping and reserve, where solar-heavy procurement creates steep daily ramps that must be absorbed by hydro dispatch, batteries, or cross-border trades. The fourth constraint is balancing and settlement: the true cost to an industrial buyer is not just the PPA strike; it is the strike plus shaping plus imbalance plus replacement volumes during curtailment or outage periods. Those add-ons can quietly add several euros per MWh to the effective delivered cost if the portfolio is not aggregated and actively managed.
This is where aggregation and virtual balancing are no longer optional features; they are the commercial enabler. A portfolio-level aggregator can combine geographically diversified wind, node-optimized solar, storage dispatch, and intraday repositioning to deliver firmed green blocks to CBAM-exposed industry. This does not eliminate physics, but it reduces forecast error, reduces imbalance costs, and converts “intermittent generation” into “deliverable procurement products.” For a platform delivering 2.0–3.0 TWh per year, even a net improvement of €3–5/MWh from aggregation, better shaping, and reduced imbalance penalties translates into €6–15 million of annual value—often the difference between a project that survives the downside and one that fails under stress.
The most important stress test is the one you flagged: a grid-upgrade slip of 12–18 months. In portfolio programs, the failure mode is not “everything delays”; it is “some nodes delay,” stranding specific tranches. If 300 MW of planned capacity is delayed by 18 months, deferred generation can easily reach 700–1,000 GWh depending on whether it is wind- or solar-heavy. At €70–90/MWh, that is €49–90 million of postponed revenue and attribute delivery, concentrated in early years when financing carry and contract milestones matter most. Equity IRR impacts typically compress by 100–200 basis points in a disciplined base case and 150–250 basis points in an aggressive upside case, especially if delays coincide with increased grid CAPEX and forced merchant exposure. Wind-heavy portfolios degrade more gracefully because wind can be geographically distributed across nodes, is less synchronized, and retains stronger capture prices even when commissioning slips.
The strategic conclusion is that Serbia’s CBAM-exposed exporters cannot treat green electricity as a generic “PPA at the lowest strike.” They need a portfolio solution that delivers dependable attribute volumes under real grid constraints. The bankable approach is a wind-anchored supply stack, solar deployed selectively where nodes are strongest, storage sized as insurance rather than decoration, and aggregation built in from the start to firm delivery and protect capture prices. Under that model, CBAM shifts from being a border tax risk to being a procurement-driven competitive differentiator. Without it, the margin erosion will not arrive as a single shock; it will arrive as procurement teams quietly repricing Serbia’s suppliers contract by contract until the volume migrates elsewhere.
Elevated by clarion.energy
