While the cost of grid power in South Africa continues to rise at double-digit-percentage rates, the cost of producing power from solar energy is falling – and AI optimisation tools are playing a key role.

That’s according to residential subscription solar service provider GoSolr’s June 2025 “Light Paper”, published late last week.

“Innovation and the growth of tech are the solutions that will grow solar use and help us work towards the promised just transition of power,” said GoSolr. “Green tech like AI and electric vehicles are driving the efficiency and uptake of renewable energy. Wheeling is making the access to renewable energy easier and micro-grids can help reduce the load on the national grid.”

South Africa experienced its worst year of load shedding in 2023. The rolling power cuts were so severe that residential rooftop installations ballooned from 983MW in March 2022 to 4.7GW by July 2023. According to GoSolr, residential installations have slowed since but are still strong, with homeowners now looking to solar more as an escape from expensive, Eskom-supplied grid power.

Energy regulator Nersa approved a 12% electricity increase for Eskom in 2025. According to GoSolr CEO Andrew Middleton, around 80% of the electricity sold by Eskom goes through municipalities. Some municipalities, like the City of Cape Town, have chosen to absorb a portion of the increase (1%) and to pass the rest onto its customers (11%). Others, including Johannesburg, have passed the full 12% onto consumers, with other costs, including a fixed usage tariff, in the pipeline.

The role of AI

“Eskom is wanting to charge more for that fixed-charge component and potentially less for the variable part. Unfortunately, the current proposal is making that fixed charge very high in places like Johannesburg, to the point where it is going to be about R1 400 if they get what they are currently proposing. That’s before you have spent R1 on electricity,” said Middleton.

He said the cost of producing electricity from a home solar installation, excluding the expensive cost of battery storage, is around 50% less than the average South African household pays for grid power. AI optimisation tools like GoSolr’s “The Brain” can drive these costs down by a further 20-40%, he added. Competitor Wetility has a similar product called “AI Mode”.

Read: Homeowners are still going solar – but for different reasons

Homeowners with solar installations already have tools, including sensors and mobile apps, that allow them to monitor and tweak their systems for optimal energy efficiency. However, a lot of manual intervention is still required, including checking the weather to see if the geyser may need to be left on for longer on a particular day, for example.

AI tools automate this process by developing predictive models based on system telemetry data, usage patterns and external sources like weather forecasts. AI systems can also predict when system components will need maintenance and give warnings if something is about to break.

“Software is emerging as the critical layer that determines whether a system merely functions, or whether it truly delivers value. From optimising energy flows and adapting to volatile grid conditions to enabling remote diagnostics and predictive maintenance, intelligent software is reshaping what solar can do and for whom,” said Wetility in a recent statement.

Users of home solar with bi-directional meters can in cities like Cape Town also feed their excess power to the grid at times when their own demand for electricity is low. This is typically during the day when the sun is shining and homes are likely to be empty. Other municipalities, including the City of Johannesburg, are working to develop feed-in tariffs to facilitate the same functionality.

Johannesburg is also exploring time-of-use tariffs that could see households and businesses pay more for electricity during peak hours. GoSolr believes reciprocal feed-in tariffs ought to be implemented as well so that homeowners who offload to the grid at peak times also earn more than when they do off-peak.

Complexity

However, this arrangement would add to the complexity of managing home solar installations because households systems would need to balance household draw against cheaper daytime grid tariffs and feeding into the grid at the most optimal times – and this is where AI tools are most useful, said Middleton.

Away from individual households, micro-grids – whether built by communities, estates or business parks – are seen as the next step in decentralising the power supply and further stabilising the national grid. But the problem of managing intermittent demand from households – sometimes they want power from the grid and at other times they opt to feed excess onto the grid – only compound further at the micro-grid level.

Read: Solar panic? The truth about SSEG, fines and municipal rules

According to Middleton, AI is going to play an even more important role in this layer, with intelligent networks predicting usage patterns and reporting to Eskom ahead of time how much power a particular micro-grid will want to draw or feed into the system. – © 2025 NewsCentral Media

Get breaking news from TechCentral on WhatsApp. Sign up here.