Analysis
The Republic of Congo's news agency ACI reported on 24 September 2026 that officials had visited the national data centre at Bacongo, Brazzaville, and the Kintélé solar site two days earlier. The centre was described as 82% complete, with operation targeted by year-end. ACI reported 19,354 photovoltaic panels, 12 MW of solar capacity and eight converters at Kintélé, with solar and the national E2C grid intended to support the centre. A site visit and a target date do not establish commissioning or measured uptime.
ACI also printed a figure of '40 megawatts per hour'. MW per hour describes a rate of change in power, not stored energy, battery power or generation output. The underlying quantity cannot be recovered from the sentence, so this analysis does not call it a 40-MWh battery. That ambiguity matters: a data centre's continuous load requires night supply and fast transfer between sources, while a solar array produces variable daytime energy. The following calculations are explicitly hypothetical sizing examples, not project estimates.
Evidence and site boundary
The 24 September ACI report provides an official site-visit account and identifies E2C grid supply alongside solar. It says the data centre was 82% complete, which is a construction-status statement. It does not specify whether the reported 12 MW is DC module rating or AC output, the electrical distance between sites, the data centre's IT load or total facility load.
An African Development Bank environmental and social document establishes an earlier national data-centre element in the Central Africa Backbone programme, but is not evidence of this facility's September electrical performance. The current analysis rests on ACI for the 2026 site facts and treats older programme documents only as background.
Power balance, not panel count
If 12 MW were available at the AC connection, output would still vary with irradiance, temperature, soiling and curtailment. Nineteen thousand panels do not establish hourly energy without module rating, inverter sizing and operating conditions. Grid import may cover the remainder, but its reliability and interconnection constraints are not quantified in the report.
Data-centre power includes IT equipment plus cooling, conversion and auxiliary systems. A power-usage-effectiveness ratio cannot be assumed from the source. Critical equipment also needs voltage and frequency quality and a fast response to grid or solar disturbances; an annual solar-energy total cannot answer an outage question.
Illustrative energy arithmetic
As a hypothetical case only, suppose the site supplies 12 MW AC at a 20% annual capacity factor. Energy would be 12 × 0.20 × 8,760 = 21,024 MWh/year. At 15% and 25%, the corresponding figures are 15,768 and 26,280 MWh/year. These are sensitivity values, not measured Congo output or an estimate of the site's true capacity factor.
A hypothetical facility drawing 2 MW continuously would need 17,520 MWh/year and 2 MW every hour. Even an annual solar total above that annual load would not prove round-the-clock coverage, because daytime surplus cannot cover nights without known storage or a sufficiently reliable grid. Distribution losses and uptime requirements further weaken a simple annual comparison.
Economic boundary
The possible value of solar is reduced grid purchases or diesel use, subject to coincidence of generation and load, delivered electricity prices, backup cost and maintenance. ACI describes a more resilient and economical model as an expectation, but publishes no plant capital cost, power tariff, fuel expense or outage record from which savings could be measured.
For a purely hypothetical 10,000 MWh/year of solar energy that actually displaces purchased electricity at a hypothetical US$0.10/kWh, gross avoided purchases equal US$1 million/year. That is 10 million kWh times US$0.10, before network, storage, maintenance or financing costs. It is not a valuation of Kintélé.
What would resolve uncertainty
A commissioning certificate, metered 15-minute solar output, data-centre IT and total-facility demand, E2C outage logs and a single-line electrical diagram would establish the real supply boundary. Storage should be reported in both MW and MWh, including usable state-of-charge limits, rather than reconstructed from ACI's ambiguous wording.
The 2026–2027 test is whether the centre opens, maintains required uptime and uses solar energy during actual demand hours. Until those observations exist, the reported hardware is a plausible component of a diversified supply plan, not proof of autonomous or cheaper data-centre operation.