Back to Articles

Is AI really the new electricity?

When Andrew Ng made that pivotal statement in 2017, he was generalizing the impact of AI on society. He was saying that AI would be as transformative as electricity was in the 19th century:

Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don’t think AI will transform in the next several years.

But for me, I’m seeing a much broader parallel when it comes to thinking about how AI should be approached by business and policy stakeholders in Africa.

Let’s look intently what AI and electricity have in common, espcecially in this era of generative AI (forgive the pun).

The Advent

Neither electricity nor AI was invented in a single moment by a single person. Just as electricity evolved from Faraday’s principles in the 1830s to Edison’s practical lighting systems in the 1880s, AI has grown from Turing’s theoretical foundations in the 1950s to today’s practical applications. The pioneers of both technologies – from Edison and Tesla to modern AI researchers like Geoffrey Hinton, Yoshua Bengio, and Yann LeCun – built upon each other’s work to create something transformative.

The Generation Parallel

In my country, we recently saw ESCOM (the Electricity Supply Corporation of Malawi) split into two entities: EGENCO for generation and ESCOM for distribution. This separation mirrors exactly what we’re seeing in the AI industry today.

Just as it wouldn’t make economic sense for every business in Malawi to generate its own electricity, it rarely makes sense for organizations to build their own large language models.

The economics are striking. While EGENCO invests millions in power stations, training a single large language model can cost upwards of $100 million. Both require massive infrastructure investments that only make sense at scale. Today’s AI “power plants” – companies like OpenAI (backed by Microsoft), Anthropic (backed by Amazon), and Google – are playing the same role that early electricity generators played a century ago.

The Distribution Parallel

The distribution layer can be just as crucial as the generation. In the electricity world, we understand this well. ESCOM doesn’t generate power anymore, but their role in distributing it efficiently to homes and businesses is absolutely critical. They primarily focus on effectively managing the grid, ensuring that power reaches everyone who needs it.

Similarly, while companies like OpenAI and Anthropic generate AI capabilities, it’s the distributors – cloud providers, API and SDK services, and hosting platforms – that make AI accessible to businesses. Hugging Face is a great example of a company that has built a distribution platform for AI models. They’ve made it easy for businesses to access and deploy state-of-the-art models without having to build them from scratch.

LangChain is another company that was early into filling the AI distribution gap, and today leads the way developers interact with AI LLMs to build applications.

Let their be Light

Every techology needs a killer app to drive adoption. For Blockchain, it was Bitcoin and Cryptocurrency in general. For Electricity, it was simply lighting! Just as electric lighting became electricity’s first “killer app,” chatbots have emerged as AI’s universal application.

Think about it: Edison’s Pearl Street Station started with just 400 customers in Manhattan, but by 1930, 70% of US homes had electric lighting. Similarly, ChatGPT reached 100 million users in just two months after its launch in late 2022. Both technologies share that magical quality of being immediately useful to almost everyone, with an interface so simple anyone can use it – flip a switch, type a question.

The Multiplier Effect

But here’s the most important parallel, and the one that most organizations miss: The true transformative change in electricity didn’t come from power plants – it came from appliances. The electric stove freed people from gathering firewood. The refrigerator transformed food preservation. The sewing machine revolutionized clothing production. The radio transformed access to information. Each of these appliances took the raw power of electricity and turned it into something that solved specific, practical problems.

Just as a new field of Electrical Engineering emerged from the electricity era, today’s AI Engineers will focus on building applinces for the AI era. Specific AI Agents that solve problems for specific use cases and communities. This is where I personally believe the real value for most organizations lies. This is where the multiplier effect of AI will be felt most strongly.

Whether it’s a chat-gpt wrapper that helps farmers answer questions about their crops in their local language or an on-edge fine-tuned Document AI model that helps local governments digitize their records, the real value of AI will come from these AI appliances, because it will take a deep understanding of the local context to build them, and make them truly useful to the people who need them.

Conclusion

Business leaders will need to think critically about where their investments fit in this AI value chain. Before you ask the question:

Why are we not building our own Large Language Models here in Africa?

We need to examine carefully at what stage we want to play. Are we asking because we simply want to participate in the AI revolution, or do we have a clear framework of thinking about AI and how we can unlock meaningful value for our organizations and communities?

As massive amount of grant money will be poured into AI in Africa over the next decade, having this framework will be crucial to ensure that we’re getting the most out of these investments.