A training run is a power contract, a land decision and a cooling loop before it is a model. Keraunos exists to hold all three at once, in a region where the energy is already built and the load has not arrived yet.
Most AI capacity in South America is assembled from four parties: a generator, a landowner, a developer and a cloud. Every schedule slip lives in the gaps between them.
We hold all four functions in one company. Origination, energy, construction management and cloud operations report to the same team, so a delivery date is something we control rather than something we forecast.
That is also why our cost of compute falls with each site. The contract template, the retrofit specification and the operating standard do not change from one region to the next. Only the building does.
We do not break ground on capacity we cannot energise. Supply and interconnection come first, every time.
Performance, uptime and cost are answered by the team that owns the racks, not escalated to a landlord.
South American jurisdiction, data residency in country, and commercial terms that hold across decades rather than election cycles.
Customers get a specification, a region and a date. If we cannot hold the date we say so before the contract, not after.
Energy and site work run on the ground in Brazil, alongside a US-incorporated contracting entity.
Each site is faster and cheaper to bring online than the last, because nothing about the method is bespoke.
Two sides of the same business: customers who need compute on a date, and partners who hold power, land or halls.
Frontier and open-weight teams running long training campaigns.
Products at production volume that need reserved throughput.
Regulated workloads needing in-country residency and a named operator.
Renewable portfolios looking for firm, high-utilisation load.
Tier III halls that can be converted to AI density.
Long-horizon capital funding campus development against contracted load.
We are hiring across energy development, data-centre engineering and cloud operations.