Power before fibre: underwriting a small-market data centre

The division reviewed nine data centre proposals in small and medium jurisdictions over the past eighteen months and committed to two, one in the Caribbean and one in the Indian Ocean. The demand case in each was similar and sound: a government data residency requirement, a financial sector that must keep records locally, and a set of businesses currently hosting several thousand kilometres away with the latency that implies. Demand was never the reason the other seven were declined. Six failed on power and one on land title.

An 8 megawatt facility needs a connection the local utility must build, and the queue for that connection is the real construction schedule. In the seven declined proposals, the median quoted connection date was 26 months after financial close, and in three cases the utility would not issue a firm date at all. A promoter who treats this as an administrative step has not read the capital plan of the utility. We do not commit before a signed connection agreement with a firm date and stated liquidated damages, and that single rule removed most of the pipeline.

Where grid supply is thin, the facility runs on its own generation more often than the model assumes. Both committed projects carry generation sized to full load with fuel storage for 96 hours and a contracted resupply arrangement tested quarterly. Fuel is a pass-through to tenants in the contracts, subject to a cap, because a facility that absorbs fuel volatility in a jurisdiction with import dependence is a commodity position wearing an infrastructure label. Cooling is the other physical constraint. In tropical conditions, cooling can take 35 to 45 per cent of total load, and a design imported from a temperate market will miss its power usage target by a wide margin.

Pre-leases decide the return. At financial close, the Caribbean project held signed commitments for 62 per cent of capacity across four tenants, with a weighted average term of seven years and the two largest being regulated financial institutions with a residency requirement. Our underwriting produces an unlevered return of 11.5 per cent on that basis and 7.1 per cent on a speculative build. The difference is the whole investment case. The division rule is two anchor tenants under signed leases before commitment, and a cap on any single tenant at 30 per cent of contracted capacity.

The operating risk that most models understate is people. A facility needs technicians who can maintain switchgear, generation and cooling in a market where perhaps three other employers need the same skills. Both projects fund a training programme with a local technical institution and carry a maintenance contract with an overseas provider for the first three years, with a defined handover. Where that transfer is not planned, the facility runs on flown-in engineers at four times the cost, and the margin the model showed disappears in the second year of operation.

These positions suit this firm for a specific reason. They require patience, local relationships and a tolerance for regulatory process, which is the same profile as the origination sectors the firm knows. They are funded from realised proceeds rather than from new external capital, and the Caribbean project was funded in part from the specialty materials realisation completed in the fourth quarter of last year. Capital earned in a regulated manufacturing niche has become infrastructure in a jurisdiction that needs it, which is the model working as intended.

Published 2023-08-15 by the Private Equity division. Research is prepared for eligible counterparties and does not constitute advice.