Power, grid and self-reliant energy

In a growing number of markets, the limiting factor on AI infrastructure is not capital, land or hardware. It is the electrical capacity available at a site and the queue to increase it, which in constrained regions runs to years.

AI compute is an unusual electrical customer: high density, high load factor, and relatively insensitive to location provided latency and jurisdiction requirements are satisfied. That combination is exactly the profile that behind-the-meter generation and flexible operation are theoretically well suited to serve, and it is why grid strategy has become an infrastructure design question rather than a procurement detail.

Whether pod-scale deployment can be sized to fit within existing building capacity often enough to be a systematic deployment-speed advantage rather than an occasional one. Power architectures suited to pod scale, including higher-voltage DC distribution and conversion efficiency treated as a thermal parameter rather than only an electrical one. Storage integrated into the pod thermal architecture rather than bolted alongside it. Whether a distributed fleet of pods across many sites is a better flexibility resource than a single campus, which CoralDC believes is likely and has not yet demonstrated.

Grid-interactive and flexible-load compute. Immersion-cooled energy storage sharing one fluid and one thermal architecture with the compute. Power electronics and conversion at pod scale. Superconducting power delivery, as a long-horizon item. On-site and behind-the-meter generation, through partners rather than as a CoralDC-delivered capability.

Whether load flexibility has monetisable value depends entirely on market structure, and most demand response programmes were not designed for this load class. In several of CoralDC's target markets there is currently no mechanism that would pay for it. CoralDC regards flexibility as strategically important and commercially unproven in its home market, and will not present it as a revenue line until a specific market mechanism has been identified.

Whether an institutional customer will accept workload interruption, even brief and automated, is a behavioural question rather than a technical one, and CoralDC does not yet have enough deployment evidence to answer it.

Utilities and system operators examining large flexible load. Energy developers structuring behind-the-meter arrangements. Storage and power electronics companies. Grid software companies.

Utility strategy and demand-side teams. Energy project developers. Power electronics and storage companies. Researchers working on flexible industrial load.