Mexico's data center PPAs are priced against a queue nobody controls
Querétaro held a handful of data centers in 2022. By 2025, Microsoft, Google, AWS, KIO and ODATA had installed or begun building more than 800 MW of hyperscale capacity there, roughly what a city of 600,000 people consumes. The Bajío grid behind that load, designed for dispersed industrial demand, now runs an interconnection queue of about 14 months, and data centers specifically are waiting 6 to 24 months for a connection point, per reporting on CFE's own congestion data. Every renewable PPA signed against that load is priced to a commercial operation date that the grid, not the generator, actually sets.
The story getting told about Mexican data centers is a demand story: hyperscalers need clean power to meet their own sustainability commitments, and that need is supposed to translate directly into new contracted megawatt-hours for whoever can build generation fast enough. It is not wrong. It is also not the constraint that matters most right now.
The demand is real. The queue is the part everyone skips
Generation interconnection risk in Mexico gets plenty of attention, and it should. What gets less attention is that a PPA between a generator and a data center has two interconnection points in it, not one. The generator has to connect to the grid. So does the building the electrons are ultimately serving. In Querétaro today, the second connection is frequently the slower one, and it belongs to a party the generator does not control, cannot expedite, and in most contract structures cannot even see the status of in real time.
That asymmetry does not show up in a standard offtake term sheet. A PPA states a commissioning date, a contract tenor and a price. It says nothing about whether the load it is written against has a breaker position to receive the power through on that date.
A signed PPA with a data center is a promise about megawatt-hours and a year. It is not a promise about whether the wire on the other end of it gets built on time.
CFE is rationing a queue, not clearing one
CFE's total 2026 budget fell 16.7% in real terms from 2025, even as the utility raised Querétaro-specific spending by 51.3%, to roughly MX$64.3 million, funding two transmission lines and a round of substation work aimed squarely at Bajío congestion. Set against 800-plus MW of hyperscale load layered onto a grid never sized for that density, that increase is real but still small relative to the gap it is meant to close. CFE has already forced connection cuts on existing industrial users to make room for new load, which is a rationing decision dressed up as a capacity upgrade. The tiebreaker in that rationing is not who has signed the largest PPA. It is queue position, and manufacturers who arrived for nearshoring in 2019 through 2022 hold earlier positions than hyperscalers signing today.
What we think follows
Developers selling generation into Mexican data center demand should underwrite commercial operation against the offtaker's own interconnection milestone, not the date printed in the PPA. A wind or solar asset finished on schedule earns nothing if the load it was contracted to serve is still sitting in a CFE queue. That means diligence on the offtaker's own connection agreement, not just its balance sheet, becomes a financing condition, which is an unusual ask of a hyperscaler and one most term sheets do not currently make.
There is a fair counterargument. Hyperscalers carry balance sheets and political weight that utilities everywhere respond to, and CFE's own 51% jump in Querétaro spending is evidence that pressure can move capex faster than an ordinary planning cycle would. That is plausible, and it may compress the queue for the largest, most visible loads specifically. It still does not tell a lender, today, which side of a 6-to-24-month range a given connection lands on.
ZFC has no assets in Mexico. Our 870+ MW pipeline sits in Peru, where Olmos Wind (approximately 135 MW, Lambayeque, RtB Q1 2028), Frontera (approximately 135 MW, Lambayeque, RtB 2028), Aurelion (300 MW bifacial solar, Moquegua, RtB 2028) and Solar II (300 MW bifacial solar, Moquegua, RtB 2029) interconnect through COES rather than a vertically integrated utility managing both sides of a bilateral contract. We watch Querétaro because it names a failure mode we underwrite against on our own generation side: a project finished on schedule is not the same asset as a project earning revenue on schedule, and the gap between those two dates belongs on a term sheet, not in a footnote.
The open question
Does CFE's Bajío transmission capex actually catch up with hyperscaler demand inside the 2026-2028 window fast enough for PPAs being signed this year to start paying on schedule? Or does Mexico end up with a wave of financially closed, generation-ready renewable projects sitting idle, waiting on a breaker position for a data center that is still waiting on one of its own? I do not think anyone modelling this today has a confident answer.
See the full active portfolio or read more Insights.