South Africa has potential to play a significant role in the global race towards artificial general intelligence (AGI) but the country’s energy challenges could impede its progress. AGI's vast computational needs will escalate electricity demands exponentially, highlighting the urgency of shifting towards renewable energy, says Mandy Hattingh, Legal Practitioner at law firm NSDV.
Narrow AI is designed for specialised tasks like natural language processing, playing chess or detecting spam. In contrast, AGI will resemble a polymath, capable of addressing complex problems across various fields. Its emergence is expected to revolutionise sectors such as healthcare, scientific research and automation.
As AI evolves, the energy required for training models like ChatGPT has increased dramatically. For instance, training ChatGPT-3 took 34 days and consumed approximately 1 248 MWh of electricity while training ChatGPT-4 consumed around 50 GWh of electricity. The shift from periodic training to continuous learning will further increase electricity consumption as AI systems update their knowledge in real time. South Africa must secure reliable, affordable and sustainable energy supply to compete in the AGI race.
While South Africa has made strides in renewable energy, several obstacles remain. These include policy inconsistencies such as the recent imposition of a 10% tariff on imported solar panels and discontinuation of tax incentives for solar installations. Such measures discourage investment in renewables at a time when incentives are crucial to accelerate green energy adoption.
To benefit from the AI revolution, South Africa must prioritise investment in renewable energy to support data centres, addressing rising energy demands and mitigating potential development setbacks. Powering data centres with renewables and storage solutions like green hydrogen or batteries can reduce greenhouse gas emissions, lower long-term costs amid proposed Eskom tariff hikes, ease grid strain, boost global competitiveness by attracting investment and potentially feed excess power back into the grid for additional revenue and energy security.
The National Data and Cloud Policy, published in May, prioritises data centres with self-sufficient energy sources. However, without adequate incentives to support renewable energy infrastructure, the policy may struggle to drive significant change. Renewable energy adoption is critical to power data centres sustainably and reduce greenhouse gas emissions while ensuring South Africa remains competitive in AI development.
Several African countries have successfully incentivised renewable energy investment. For example, Kenya exempts solar and wind energy equipment from VAT, Ghana imposes no VAT on imported solar panels and Botswana has waived import duties on solar equipment. South Africa can follow suit by offering tax breaks and reducing tariffs on renewable energy technology to stimulate growth in the green energy sector.
Powering data centres with renewables would not only reduce operating costs in light of rising electricity tariffs but also alleviate strain on the national grid. Co-locating renewable energy facilities with data centres would minimise energy losses and support broader energy security by supplying excess power to the community. These efforts would enhance the country’s attractiveness to foreign investors, especially as the global economy increasingly favours sustainable energy practices.
Encouraging renewable energy investment requires a combination of “carrot and stick” approaches. While government’s current stance leans towards restrictive measures, more supportive policies, such as tax breaks for renewable energy hardware, would incentivise adoption. These strategies would not only support AI and AGI development but also drive economic growth through green jobs.
