Chile's data-center boom is solving the wrong power problem
Chile's data-center load is projected to climb from roughly 325 MW in 2025 to about 1,207 MW by 2030, per CEN estimates reported by El Desconcierto, almost all of it landing in the Metropolitan region around Santiago. The pitch circulating in development and lending circles is that this demand finally gives Chile's curtailed solar a buyer. It mostly won't, and the reason has nothing to do with how much power exists.
Chile curtailed roughly 6,084 GWh of solar and wind in 2025, per Acera figures drawn from Coordinador Eléctrico Nacional dispatch data, concentrated in the handful of clear-sky midday hours when northern solar floods a grid that cannot move it south fast enough. A data center does not run at midday. It runs at every hour, pulling a flat load with a realized capacity factor well above 90%, and it does not care what the sun is doing when it needs power.
Why flat load and peaked supply don't just net out
A hyperscaler negotiating offtake is not buying generation. It is buying a number of megawatts it can enter into its own capacity plan, available whenever its racks draw it, with financial penalties attached if the counterparty can't deliver on an off hour. Solar delivers abundance at noon and nothing at night. Matching the two means somebody contracts to fill the gap, either with storage duration long enough to shift noon output into the evening and overnight hours, or with a second, dispatchable source standing behind the renewable one.
That is shaping, and it is a cost, not a footnote. Chile's own capacity-payment reform, DS N°1 of 2026, pays full credit only above five hours of battery duration, and most of the storage actually under construction today is financed at two to four. A developer trying to shape solar into a flat data-center load with under-duration storage is buying the gap twice: once in the batteries it builds, and again on the spot market whenever those batteries run dry before the load does.
Shaping cost is the line item that decides whether this demand rescues curtailed solar or just builds a second grid problem next to the first one.
Debt sized against the wrong curve
Lenders underwriting a renewable PPA against a hyperscaler's credit are used to sizing debt off a P50 generation curve and an investment-grade counterparty rating. A flat-load contract changes what sits on the other side of that curve. The exposure isn't whether the plant generates its expected GWh. It's whether the plant, plus whatever firming capacity sits behind it, can deliver a flat number at 3am in July, something a standalone solar asset was never built to do. Treat that gap as a modeling footnote and it shows up later as a draw on a liquidity facility nobody sized for it.
The queue gets more crowded, not less
Here is the part that should worry anyone counting on this demand to fix curtailment. Firm capacity built to serve a flat data-center load needs its own interconnection point, usually at the same Metropolitana substations already backed up behind roughly 30 GW of battery storage filings against 2.3 GW actually running, a queue problem we've written about before. New firm capacity competing for those same nodes does not relieve the congestion producing the curtailment. It adds another claimant to it.
What we think follows
Chile moved fast on the easy half of this problem. Its Sectoral Authorizations Framework Law reportedly cut data-center permitting from up to 120 days to a maximum of 25. Nothing has moved anywhere near that fast on the power side. My own read is that this demand mostly builds a second, parallel capacity stack, firm, storage- or thermal-backed, sited specifically to serve hyperscalers, rather than finding a way to monetize the solar that's already stranded. That is not a bad outcome for the country's power balance. It is a different outcome than the one most pitch decks describe.
There's a real counter to this. If enough five-hour-plus storage gets built chasing data-center contracts, some of that duration inevitably ends up shifting solar into evening hours too, cutting curtailment as a side effect of chasing an entirely different customer. That's plausible. It's also a few years and a lot of capital away from where the queue sits today.
Our own pipeline sits in Peru, not Chile, where flat-load corporate offtake for digital infrastructure hasn't arrived at anything like this scale yet. The question will land the moment it does. Olmos Wind, the most advanced project in our 870+ MW pipeline at roughly 135 MW in Lambayeque, carries battery storage integrated into the design rather than bolted on after Ready-to-Build, targeted for the first quarter of 2028, which is the kind of decision this exact problem rewards.
The open question
What I don't know is whether CNE and CEN route new hyperscale load through firm contracts sitting behind the same congested Metropolitana nodes as the curtailed solar, worsening a queue that's already more than ten times oversubscribed, or force a transmission build specifically for load growth that stays separate from the renewable interconnection process. Nobody I've read has committed to either path yet.
See the full active portfolio or read more Insights.